* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
--source include/not_embedded.inc
|
|
|
|
# Don't test this under valgrind, memory leaks will occur as we crash
|
|
|
|
--source include/not_valgrind.inc
|
|
|
|
# Binary must be compiled with debug for crash to occur
|
|
|
|
--source include/have_debug.inc
|
|
|
|
--source include/have_maria.inc
|
|
|
|
|
2007-12-18 22:55:49 +01:00
|
|
|
set global maria_log_file_size=4294967295;
|
2009-02-15 11:58:34 +01:00
|
|
|
let $MARIA_LOG=../../tmp;
|
2007-12-13 11:26:55 +01:00
|
|
|
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
--disable_warnings
|
|
|
|
drop database if exists mysqltest;
|
|
|
|
--enable_warnings
|
|
|
|
create database mysqltest;
|
WL#4374 "Maria - force start if Recovery fails multiple times"
http://forge.mysql.com/worklog/task.php?id=4374
new option --maria-force-start-after-recovery-failures=N; number of consecutive recovery failures (failures
of log reading or recovery processing, anything in [translog_init(),maria_recovery_from_log()])
is stored in the control file; if at a Maria start they are more than N, logs are removed. This is for automated
systems which have to run whatever happens. As tables risk staying corrupted, --maria-recover should also
be used on them: this revision makes maria-recover work (it was disabled).
Fixed bug in translog_is_log_files(). translog_init() now prints message to error log if failed.
Removed \0 in the output of SHOW ENGINE MARIA LOGS; removed hard-coded engine name there.
KNOWN_BUGS.txt:
As option --maria-force-start-after-recovery-failures is added, it corresponds to the wish "we should fix that if this happens etc".
LOAD INDEX is not ignored since a few weeks. Listed concurrency bugs have been fixed some time ago.
Recovery of fulltext and GIS indexes works since a few weeks.
mysql-test/include/maria_make_snapshot.inc:
configurable prefix in table's name (so far 't' or 't_corrupted')
mysql-test/include/maria_make_snapshot_for_comparison.inc:
configurable prefix in table's name (so far 't' or 't_corrupted')
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
configurable prefix in table's name (so far 't' or 't_corrupted')
mysql-test/include/maria_verify_recovery.inc:
configurable prefix in table's name (so far 't' or 't_corrupted')
mysql-test/lib/mtr_report.pl:
new test maria-recover.test generates expected corruption warnings in the error log. maria-recovery.test's corrupted table is renamed to t_corrupted1 instead of t1.
mysql-test/r/maria-preload.result:
result update. maria_pagecache_read* values are similar to the previous version of this file, though a bit bigger
because using the information_schema and the join leads to some internal maria temp table being used, and thus some
blocks of it being read.
mysql-test/r/maria-purge.result:
engine's name in SHOW ENGINE MARIA LOGS changed.
mysql-test/r/maria-recover.result:
result for new test. We see corruption messages at first SELECT and then none at second SELECT, expected.
mysql-test/r/maria-recovery.result:
result update
mysql-test/r/maria.result:
new variables show up
mysql-test/t/disabled.def:
BUG#34911 is not fixed but the test had been made independent of the bug (workaround). A new bug (crash) has popped recently, so it has to stay
disabled (BUG#35107).
mysql-test/t/maria-preload.test:
Work around BUG#34911 "FLUSH STATUS doesn't flush what it should":
compute differences in status variables before and after relevant queries
mysql-test/t/maria-recover-master.opt:
test --maria-recover
mysql-test/t/maria-recover.test:
Test of the --maria-recover option (build a corrupted table and see if it is auto-repaired)
mysql-test/t/maria-recovery-big.test:
update for new API of include/maria*.inc
mysql-test/t/maria-recovery-bitmap.test:
update for new API of include/maria*.inc
mysql-test/t/maria-recovery.test:
update for new API of include/maria*.inc. Corrupted table t1 renamed to t_corrupted1, so that mtr_report.pl
does not blindly remove all corruption messages for t1 which is
a common name.
storage/maria/ha_maria.cc:
Enabling maria-recover.
Adding option and global variable --maria_force_start_after_recovery_failures: ha_maria_init()
calls mark_recovery_start() and mark_recovery_success() to keep track of failed consecutive recoveries
and remove logs if needed.
Removed \0 in the output of SHOW ENGINE MARIA LOGS; removed hard-coded engine name there.
storage/maria/ma_checkpoint.c:
new prototype
storage/maria/ma_control_file.c:
Storing in one byte in the control file, the number of consecutive recovery failures.
storage/maria/ma_control_file.h:
new prototype
storage/maria/ma_init.c:
new prototype
storage/maria/ma_locking.c:
Need to update open_count on disk at first write and close for transactional tables, like we already did for
non-transactional tables, otherwise we cannot notice that the table is dubious.
storage/maria/ma_loghandler.c:
translog_is_log_files() is made more generic to serve either to search or to delete logs (the latter is
for --maria-force-start-after-recovery-failures). It also had a bug (always returned FALSE).
storage/maria/ma_loghandler.h:
export function because ha_maria::mark_recovery_start() needs it
storage/maria/ma_recovery.c:
changing name of maria_recover() to distinguish from the maria-recover option.
storage/maria/ma_recovery.h:
changing name of maria_recover() to distinguish from the maria-recover option.
storage/maria/ma_test_force_start.pl:
Test of --maria-force-start-after-recovery-failures (and also, to be realistic, of --maria-recover).
This is standalone because mysql-test-run does not support testing that multiple mysqld restarts expectedly failed.
I'll have to run it on my machine and also on a Windows machine.
storage/maria/unittest/ma_control_file-t.c:
adding recovery_failures to the test
storage/maria/unittest/ma_test_loghandler_multigroup-t.c:
fix for compiler warning (unused variable in non-debug build)
2008-06-02 22:53:25 +02:00
|
|
|
let $mms_tname=t;
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
|
|
|
|
# Include scripts can perform SQL. For it to not influence the main test
|
|
|
|
# they use a separate connection. This way if they use a DDL it would
|
|
|
|
# not autocommit in the main test.
|
2008-07-01 22:47:09 +02:00
|
|
|
connect (admin, localhost, root,,mysqltest,,);
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
--enable_reconnect
|
|
|
|
|
|
|
|
connection default;
|
|
|
|
use mysqltest;
|
|
|
|
--enable_reconnect
|
|
|
|
|
|
|
|
# A sample test
|
|
|
|
-- source include/maria_empty_logs.inc
|
|
|
|
let $mms_tables=1;
|
|
|
|
create table t1 (a varchar(1000)) engine=maria;
|
|
|
|
|
2007-11-13 18:49:00 +01:00
|
|
|
--echo * TEST of REDO: see if recovery can reconstruct if we give it an old table
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
|
|
|
|
-- source include/maria_make_snapshot_for_feeding_recovery.inc
|
|
|
|
# Your committed statements here, which we expect to
|
|
|
|
# be reconstructed from the log
|
|
|
|
insert into t1 values ("00000000");
|
|
|
|
-- source include/maria_make_snapshot_for_comparison.inc
|
|
|
|
# we want recovery to run on the first snapshot made above
|
|
|
|
let $mvr_restore_old_snapshot=1;
|
2007-11-13 18:49:00 +01:00
|
|
|
# As we did only committed work, we test REDO applying, which could
|
|
|
|
# produce a physically identical table.
|
|
|
|
let $mms_compare_physically=1;
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
let $mvr_debug_option="+d,maria_flush_whole_log,maria_crash";
|
WL#3072 Maria recovery:
fix for bug: if a crash happened right after writing a REDO like this:
REDO - UNDO - REDO*, then recovery would ignore the last REDO* (ok),
rollback: REDO - UNDO - REDO* - REDO - CLR, and a next recovery would
thus execute REDO* instead of skipping it again. Recovery now logs
LOGREC_INCOMPLETE_GROUP when it meets REDO* for the first time,
to draw a boundary and ensure it is always skipped. Tested by hand.
Note: ma_test_all fails "maria_chk: error: Key 1 - Found too many records"
not due to this patch (failed before).
BitKeeper/triggers/post-commit:
no truncation of the commit mail, or how to review patches?
mysql-test/include/maria_verify_recovery.inc:
let caller choose the statement used to crash (sometimes we
want the crash to happen at special places)
mysql-test/t/maria-recovery.test:
user of maria_verify_recovery.inc now specifies statement which the
script should use for crashing.
storage/maria/ma_bitmap.c:
it's easier to search for all places using functions from the bitmap
module (like in ma_blockrec.c) if those exported functions all start
with "_ma_bitmap": renaming some of them.
Assertion that when we read a bitmap page, overwriting bitmap->map,
we are not losing information (i.e. bitmap->changed is false).
storage/maria/ma_blockrec.c:
update to new names. Adding code (disabled, protected by a #ifdef)
that I use to test certain crash scenarios (more to come).
storage/maria/ma_blockrec.h:
update to new names
storage/maria/ma_checkpoint.c:
update to new names
storage/maria/ma_extra.c:
update to new names
storage/maria/ma_loghandler.c:
new LOGREC_INCOMPLETE_GROUP
storage/maria/ma_loghandler.h:
new LOGREC_INCOMPLETE_GROUP
storage/maria/ma_recovery.c:
When at the end of the REDO phase we have identified some transactions
with incomplete REDO groups (REDOs without an UNDO or CLR_END),
for each of them we log LOGREC_INCOMPLETE_GROUP. This way, the
upcoming UNDO phase can write more records for such transaction,
a future recovery won't pair the incomplete group with the
CLR_END (as there is LOGREC_INCOMPLETE_GROUP to draw a boundary).
2007-12-10 23:26:53 +01:00
|
|
|
let $mvr_crash_statement= set global maria_checkpoint_interval=1;
|
2007-11-13 18:49:00 +01:00
|
|
|
# the script below will trigger recovery and compare checksums
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
-- source include/maria_verify_recovery.inc
|
2007-11-13 18:49:00 +01:00
|
|
|
let $mms_compare_physically=0;
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
# so a SELECT like this is pure visual effect, brings nothing.
|
|
|
|
select * from t1;
|
|
|
|
|
2007-11-13 18:49:00 +01:00
|
|
|
--echo * TEST of REDO+UNDO: normal recovery test (no moving tables under its feet)
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
|
|
|
|
# different types of crash => a loop; here are loop control variables
|
|
|
|
let $crash_no_flush=1;
|
|
|
|
let $crash_flush_whole_page_cache=0;
|
|
|
|
let $crash_flush_states=0;
|
|
|
|
let $crash_flush_whole_log=0;
|
|
|
|
let $crash_loop=1;
|
|
|
|
|
|
|
|
# we want recovery to use the tables as they were at time of crash
|
|
|
|
let $mvr_restore_old_snapshot=0;
|
2007-11-13 18:49:00 +01:00
|
|
|
# UNDO phase prevents physical comparison, normally,
|
|
|
|
# so we'll only use checksums to compare.
|
|
|
|
let $mms_compare_physically=0;
|
WL#3072 Maria recovery:
fix for bug: if a crash happened right after writing a REDO like this:
REDO - UNDO - REDO*, then recovery would ignore the last REDO* (ok),
rollback: REDO - UNDO - REDO* - REDO - CLR, and a next recovery would
thus execute REDO* instead of skipping it again. Recovery now logs
LOGREC_INCOMPLETE_GROUP when it meets REDO* for the first time,
to draw a boundary and ensure it is always skipped. Tested by hand.
Note: ma_test_all fails "maria_chk: error: Key 1 - Found too many records"
not due to this patch (failed before).
BitKeeper/triggers/post-commit:
no truncation of the commit mail, or how to review patches?
mysql-test/include/maria_verify_recovery.inc:
let caller choose the statement used to crash (sometimes we
want the crash to happen at special places)
mysql-test/t/maria-recovery.test:
user of maria_verify_recovery.inc now specifies statement which the
script should use for crashing.
storage/maria/ma_bitmap.c:
it's easier to search for all places using functions from the bitmap
module (like in ma_blockrec.c) if those exported functions all start
with "_ma_bitmap": renaming some of them.
Assertion that when we read a bitmap page, overwriting bitmap->map,
we are not losing information (i.e. bitmap->changed is false).
storage/maria/ma_blockrec.c:
update to new names. Adding code (disabled, protected by a #ifdef)
that I use to test certain crash scenarios (more to come).
storage/maria/ma_blockrec.h:
update to new names
storage/maria/ma_checkpoint.c:
update to new names
storage/maria/ma_extra.c:
update to new names
storage/maria/ma_loghandler.c:
new LOGREC_INCOMPLETE_GROUP
storage/maria/ma_loghandler.h:
new LOGREC_INCOMPLETE_GROUP
storage/maria/ma_recovery.c:
When at the end of the REDO phase we have identified some transactions
with incomplete REDO groups (REDOs without an UNDO or CLR_END),
for each of them we log LOGREC_INCOMPLETE_GROUP. This way, the
upcoming UNDO phase can write more records for such transaction,
a future recovery won't pair the incomplete group with the
CLR_END (as there is LOGREC_INCOMPLETE_GROUP to draw a boundary).
2007-12-10 23:26:53 +01:00
|
|
|
let $mvr_crash_statement= set global maria_checkpoint_interval=1;
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
|
|
|
|
# Note that we don't remove logs between iterations. Test is
|
|
|
|
# cumulative (each new recovery processes more log records than the previous).
|
|
|
|
|
|
|
|
while ($crash_loop)
|
|
|
|
{
|
|
|
|
if ($crash_flush_whole_log)
|
|
|
|
{
|
|
|
|
let $mvr_debug_option="+d,maria_flush_whole_log,maria_crash";
|
|
|
|
# set up what next iteration should do:
|
|
|
|
let $crash_flush_whole_log=0;
|
|
|
|
let $crash_loop=0;
|
|
|
|
}
|
|
|
|
if ($crash_flush_states)
|
|
|
|
{
|
|
|
|
let $mvr_debug_option="+d,maria_flush_states,maria_flush_whole_log,maria_crash";
|
|
|
|
let $crash_flush_states=0;
|
|
|
|
let $crash_flush_whole_log=1;
|
|
|
|
}
|
|
|
|
if ($crash_flush_whole_page_cache)
|
|
|
|
{
|
|
|
|
let $mvr_debug_option="+d,maria_flush_whole_page_cache,maria_crash";
|
|
|
|
let $crash_flush_whole_page_cache=0;
|
|
|
|
let $crash_flush_states=1;
|
|
|
|
}
|
|
|
|
if ($crash_no_flush)
|
|
|
|
{
|
|
|
|
let $mvr_debug_option="+d,maria_crash";
|
|
|
|
let $crash_no_flush=0;
|
|
|
|
let $crash_flush_whole_page_cache=1;
|
|
|
|
}
|
|
|
|
# Your committed statements here
|
|
|
|
insert into t1 values ("00000000");
|
|
|
|
-- source include/maria_make_snapshot_for_comparison.inc
|
|
|
|
# Your statements which we expect to be rolled back
|
|
|
|
lock tables t1 write;
|
|
|
|
insert into t1 values ("aaaaaaaaa");
|
|
|
|
-- source include/maria_verify_recovery.inc
|
|
|
|
select * from t1;
|
|
|
|
}
|
|
|
|
|
|
|
|
drop table t1;
|
|
|
|
|
|
|
|
# what did we compare above:
|
|
|
|
# - checksum: tells that the tables contain the same amount of rows
|
|
|
|
# and same data in rows
|
|
|
|
# - index: no, neither state nor pages were compared
|
|
|
|
# - bitmap pages: the REPAIR QUICK done above very probably checks
|
|
|
|
# that bitmap reflects page occupation; do we need to do physical
|
|
|
|
# compare?
|
|
|
|
# - page LSN: not compared; we should compare that page's LSN in new
|
|
|
|
# table is >= page's LSN in old table (it can be >, due to UNDO phase)
|
|
|
|
# we had a bug where new page's LSN was 0... todo.
|
|
|
|
|
|
|
|
#
|
2007-11-13 18:49:00 +01:00
|
|
|
# Test for this bug: an UPDATE purges and rewrites a tail page, and
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
# recovery applied the purge, stamped page with UNDO's LSN, thus
|
|
|
|
# the rewrite was ignored.
|
|
|
|
#
|
|
|
|
|
WL#3072 - Maria recovery.
* fix for bitmap vs checkpoint bug which could lead to corrupted
tables in case of crashes at certain moments: a bitmap could be flushed
to disk even though it was inconsistent with the log (it could be
flushed before REDO-UNDO are written to the log). One bug remains, need
code from others. Tests added. Fix is to pin unflushable bitmap pages,
and let checkpoint wait for them to be flushable.
* fix for long_trid!=0 assertion failure at Recovery.
* less useless wakeups in the background flush|checkpoint thread.
* store global_trid_generator in checkpoint record.
mysql-test/r/maria-recovery.result:
result update
mysql-test/t/maria-recovery.test:
make it easier to locate subtests
storage/maria/ma_bitmap.c:
When we send a bitmap to the pagecache, if this bitmap is not in a
flushable state we keep it pinned and add it to a list, it will be
unpinned when the bitmap is flushable again.
A new function _ma_bitmap_flush_all() used by checkpoint.
A new function _ma_bitmap_flushable() used by block format to signal
when it starts modifying a bitmap and when it is done with it.
storage/maria/ma_blockrec.c:
When starting a row operation (insert/update/delete), mark that
the bitmap is not flushable (because for example INSERT is going
to over-allocate in the bitmap to prevent other threads from using
our data pages). If a checkpoint comes at this moment it will wait
for the bitmap to be flushable before flushing it.
When the operation ends, bitmap becomes flushable again; that
transition is done under the bitmap's mutex (needed for correct
synchro with a concurrent checkpoint); but for INSERT/UPDATE this
happens inside _ma_bitmap_release_unused() at a place where it already
has the mutex, so the only penalty (mutex adding) is in DELETE and UNDO
of INSERT. In case of errors after setting the bitmap unflushable,
we must always set it back to flushable or checkpoint would block.
Debug possibilities to force a sleep while the bitmap is over-allocated.
In case of error in get_head_or_tail() in allocate_and_write_block_record(),
we still need to unpin all pages.
Bugfix: _ma_apply_redo_insert_row_blobs() produced wrong
data_file_length.
storage/maria/ma_blockrec.h:
new bitmap calls.
storage/maria/ma_checkpoint.c:
filter_flush_indirect not needed anymore (flushing bitmap
pages happens in _ma_bitmap_flush_all() now). So
st_filter_param::is_data_file|pages_covered_by_bitmap not needed.
Other filter_flush* don't need to flush bitmap anymore.
Add debug possibility to flush all bitmap pages outside of a checkpoint,
to simulate pagecache LRU eviction.
When the background flush/checkpoint thread notices it has nothing
to flush, it now sleeps directly until the next potential checkpoint
moment instead of waking up every second.
When in checkpoint we decide to not store a table in the checkpoint record
(because it has logged no writes for example), we can also skip flushing
this table.
storage/maria/ma_commit.c:
comment is out-of-date
storage/maria/ma_key_recover.c:
comment fix
storage/maria/ma_loghandler.c:
comment is out-of-date
storage/maria/ma_open.c:
comment is out-of-date
storage/maria/ma_pagecache.c:
comment for bug to fix. And we don't take checkpoints at end of REDO
phase yet so can trust block->type.
storage/maria/ma_recovery.c:
Comments. Now-unneeded code for incomplete REDO-UNDO groups removed.
When we forget about an old transaction we must really forget
about it with bzero() (fixes the "long_trid!=0 assertion" recovery
bug). When we delete a row with maria_delete() we turn on
STATE_NOT_OPTIMIZED_ROWS so we do the same when we see a CLR_END
for an UNDO_ROW_INSERT or when we execute an UNDO_ROW_INSERT (in both
cases a row was deleted). Pick up max_long_trid from the checkpoint record.
storage/maria/maria_chk.c:
comment
storage/maria/maria_def.h:
MARIA_FILE_BITMAP gets new members: 'flushable', 'bitmap_cond' and
'pinned_pages'.
storage/maria/trnman.c:
I used to think that recovery only needs to know the maximum TrID
of the lists of active and committed transactions. But no, sometimes
both lists can even be empty and their TrID should not be reused.
So Checkpoint now saves global_trid_generator in the checkpoint record.
storage/maria/trnman_public.h:
macros to read/store a TrID
mysql-test/r/maria-recovery-bitmap.result:
result is ok. Without the code fix, we would get a corruption message
about the bitmap page in CHECK TABLE EXTENDED.
mysql-test/t/maria-recovery-bitmap-master.opt:
usual when we crash mysqld in tests
mysql-test/t/maria-recovery-bitmap.test:
test of recovery problems specific of the bitmap pages.
2007-12-14 16:14:12 +01:00
|
|
|
--echo * TEST of two REDOs for same page in one REDO group
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
-- source include/maria_empty_logs.inc
|
|
|
|
let $mms_tables=1;
|
|
|
|
CREATE TABLE t1 (
|
|
|
|
i int,
|
|
|
|
b blob default NULL,
|
|
|
|
c varchar(6000) default NULL
|
|
|
|
) ENGINE=MARIA CHECKSUM=1;
|
|
|
|
-- source include/maria_make_snapshot_for_feeding_recovery.inc
|
|
|
|
INSERT INTO t1 VALUES (1, REPEAT('a', 5000), REPEAT('b', 5000));
|
|
|
|
UPDATE t1 SET i=3, b=CONCAT(b,'c') WHERE i=1;
|
|
|
|
SELECT LENGTH(b) FROM t1 WHERE i=3;
|
|
|
|
-- source include/maria_make_snapshot_for_comparison.inc
|
|
|
|
# we want recovery to run on the first snapshot made above
|
|
|
|
let $mvr_restore_old_snapshot=1;
|
2007-11-13 18:49:00 +01:00
|
|
|
let $mms_compare_physically=0;
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
let $mvr_debug_option="+d,maria_flush_whole_log,maria_crash";
|
WL#3072 Maria recovery:
fix for bug: if a crash happened right after writing a REDO like this:
REDO - UNDO - REDO*, then recovery would ignore the last REDO* (ok),
rollback: REDO - UNDO - REDO* - REDO - CLR, and a next recovery would
thus execute REDO* instead of skipping it again. Recovery now logs
LOGREC_INCOMPLETE_GROUP when it meets REDO* for the first time,
to draw a boundary and ensure it is always skipped. Tested by hand.
Note: ma_test_all fails "maria_chk: error: Key 1 - Found too many records"
not due to this patch (failed before).
BitKeeper/triggers/post-commit:
no truncation of the commit mail, or how to review patches?
mysql-test/include/maria_verify_recovery.inc:
let caller choose the statement used to crash (sometimes we
want the crash to happen at special places)
mysql-test/t/maria-recovery.test:
user of maria_verify_recovery.inc now specifies statement which the
script should use for crashing.
storage/maria/ma_bitmap.c:
it's easier to search for all places using functions from the bitmap
module (like in ma_blockrec.c) if those exported functions all start
with "_ma_bitmap": renaming some of them.
Assertion that when we read a bitmap page, overwriting bitmap->map,
we are not losing information (i.e. bitmap->changed is false).
storage/maria/ma_blockrec.c:
update to new names. Adding code (disabled, protected by a #ifdef)
that I use to test certain crash scenarios (more to come).
storage/maria/ma_blockrec.h:
update to new names
storage/maria/ma_checkpoint.c:
update to new names
storage/maria/ma_extra.c:
update to new names
storage/maria/ma_loghandler.c:
new LOGREC_INCOMPLETE_GROUP
storage/maria/ma_loghandler.h:
new LOGREC_INCOMPLETE_GROUP
storage/maria/ma_recovery.c:
When at the end of the REDO phase we have identified some transactions
with incomplete REDO groups (REDOs without an UNDO or CLR_END),
for each of them we log LOGREC_INCOMPLETE_GROUP. This way, the
upcoming UNDO phase can write more records for such transaction,
a future recovery won't pair the incomplete group with the
CLR_END (as there is LOGREC_INCOMPLETE_GROUP to draw a boundary).
2007-12-10 23:26:53 +01:00
|
|
|
let $mvr_crash_statement= set global maria_checkpoint_interval=1;
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
-- source include/maria_verify_recovery.inc
|
|
|
|
SELECT LENGTH(b) FROM t1 WHERE i=3;
|
|
|
|
drop table t1;
|
|
|
|
|
WL#3072 - Maria Recovery: recovery of state.auto_increment.
When we log UNDO_KEY_INSERT for an auto_inc key, we update
state.auto_increment (not anymore at the end of maria_write() except
if this is a non-transactional table). When Recovery sees UNDO_KEY_INSERT
in the REDO phase, it reads the auto_inc value from it and updates
state.auto_increment.
mysql-test/r/maria-recovery.result:
Without the code fix, there would be in CHECK TABLE:
"Auto-increment value: 0 is smaller than max used value: 3"
and no AUTO_INCREMENT= clause in SHOW CREATE TABLE.
mysql-test/t/maria-recovery.test:
Test of recovery of state.auto_increment: from an old table,
does the replaying of the log set state.auto_increment to
what it should be.
storage/maria/ma_check.c:
new way of calling ma_retrieve_auto_increment(): pass key
storage/maria/ma_key.c:
ma_retrieve_auto_increment() now operates directly with a pointer
to the key and not on the record.
storage/maria/ma_key_recover.c:
dedicated write_hook_for_undo_key_insert(): sets state.auto_increment
under log's mutex.
storage/maria/ma_key_recover.h:
Dedicated hook for UNDO_KEY_INSERT, to set state.auto_increment.
Such hook needs a new member st_msg_write_hook_for_undo_key::auto_increment,
which contains the auto_increment value inserted.
storage/maria/ma_loghandler.c:
UNDO_KEY_INSERT gets a dedicated write_hook, to set auto_increment.
storage/maria/ma_recovery.c:
When in the REDO phase we see UNDO_KEY_INSERT: if the state is older
than this record, and the key is the auto_increment one, read
the key's value from the log record and update state.auto_increment.
storage/maria/ma_test_all.sh:
use $maria_path to be able to run from /dev/shm (faster)
storage/maria/ma_update.c:
bool is more of C++, using my_bool.
If table is transactional, state.auto_increment is already updated
in write_hook_for_undo_key_insert().
storage/maria/ma_write.c:
If table is transactional, state.auto_increment is not updated at
the end of maria_write() but rather in write_hook_for_undo_key_insert()
(under log's mutex, so that a concurrent checkpoint does not read
state.auto_increment while it is changing - corrupted).
_ma_ck_write_btree_with_log() extracts the auto_increment value
from the key, puts it into msg.auto_increment, and this is passed
to write_hook_for_undo_key_insert().
storage/maria/maria_def.h:
change of prototype of ma_retrieve_auto_increment()
storage/maria/maria_read_log.c:
use default log file size. Use separate page caches for table
and logs (needed if maria_block_size!=TRANSLOG_PAGE_SIZE).
2007-12-12 22:33:36 +01:00
|
|
|
# Test that INSERT's effect on auto-increment is recovered
|
WL#3072 - Maria recovery.
* fix for bitmap vs checkpoint bug which could lead to corrupted
tables in case of crashes at certain moments: a bitmap could be flushed
to disk even though it was inconsistent with the log (it could be
flushed before REDO-UNDO are written to the log). One bug remains, need
code from others. Tests added. Fix is to pin unflushable bitmap pages,
and let checkpoint wait for them to be flushable.
* fix for long_trid!=0 assertion failure at Recovery.
* less useless wakeups in the background flush|checkpoint thread.
* store global_trid_generator in checkpoint record.
mysql-test/r/maria-recovery.result:
result update
mysql-test/t/maria-recovery.test:
make it easier to locate subtests
storage/maria/ma_bitmap.c:
When we send a bitmap to the pagecache, if this bitmap is not in a
flushable state we keep it pinned and add it to a list, it will be
unpinned when the bitmap is flushable again.
A new function _ma_bitmap_flush_all() used by checkpoint.
A new function _ma_bitmap_flushable() used by block format to signal
when it starts modifying a bitmap and when it is done with it.
storage/maria/ma_blockrec.c:
When starting a row operation (insert/update/delete), mark that
the bitmap is not flushable (because for example INSERT is going
to over-allocate in the bitmap to prevent other threads from using
our data pages). If a checkpoint comes at this moment it will wait
for the bitmap to be flushable before flushing it.
When the operation ends, bitmap becomes flushable again; that
transition is done under the bitmap's mutex (needed for correct
synchro with a concurrent checkpoint); but for INSERT/UPDATE this
happens inside _ma_bitmap_release_unused() at a place where it already
has the mutex, so the only penalty (mutex adding) is in DELETE and UNDO
of INSERT. In case of errors after setting the bitmap unflushable,
we must always set it back to flushable or checkpoint would block.
Debug possibilities to force a sleep while the bitmap is over-allocated.
In case of error in get_head_or_tail() in allocate_and_write_block_record(),
we still need to unpin all pages.
Bugfix: _ma_apply_redo_insert_row_blobs() produced wrong
data_file_length.
storage/maria/ma_blockrec.h:
new bitmap calls.
storage/maria/ma_checkpoint.c:
filter_flush_indirect not needed anymore (flushing bitmap
pages happens in _ma_bitmap_flush_all() now). So
st_filter_param::is_data_file|pages_covered_by_bitmap not needed.
Other filter_flush* don't need to flush bitmap anymore.
Add debug possibility to flush all bitmap pages outside of a checkpoint,
to simulate pagecache LRU eviction.
When the background flush/checkpoint thread notices it has nothing
to flush, it now sleeps directly until the next potential checkpoint
moment instead of waking up every second.
When in checkpoint we decide to not store a table in the checkpoint record
(because it has logged no writes for example), we can also skip flushing
this table.
storage/maria/ma_commit.c:
comment is out-of-date
storage/maria/ma_key_recover.c:
comment fix
storage/maria/ma_loghandler.c:
comment is out-of-date
storage/maria/ma_open.c:
comment is out-of-date
storage/maria/ma_pagecache.c:
comment for bug to fix. And we don't take checkpoints at end of REDO
phase yet so can trust block->type.
storage/maria/ma_recovery.c:
Comments. Now-unneeded code for incomplete REDO-UNDO groups removed.
When we forget about an old transaction we must really forget
about it with bzero() (fixes the "long_trid!=0 assertion" recovery
bug). When we delete a row with maria_delete() we turn on
STATE_NOT_OPTIMIZED_ROWS so we do the same when we see a CLR_END
for an UNDO_ROW_INSERT or when we execute an UNDO_ROW_INSERT (in both
cases a row was deleted). Pick up max_long_trid from the checkpoint record.
storage/maria/maria_chk.c:
comment
storage/maria/maria_def.h:
MARIA_FILE_BITMAP gets new members: 'flushable', 'bitmap_cond' and
'pinned_pages'.
storage/maria/trnman.c:
I used to think that recovery only needs to know the maximum TrID
of the lists of active and committed transactions. But no, sometimes
both lists can even be empty and their TrID should not be reused.
So Checkpoint now saves global_trid_generator in the checkpoint record.
storage/maria/trnman_public.h:
macros to read/store a TrID
mysql-test/r/maria-recovery-bitmap.result:
result is ok. Without the code fix, we would get a corruption message
about the bitmap page in CHECK TABLE EXTENDED.
mysql-test/t/maria-recovery-bitmap-master.opt:
usual when we crash mysqld in tests
mysql-test/t/maria-recovery-bitmap.test:
test of recovery problems specific of the bitmap pages.
2007-12-14 16:14:12 +01:00
|
|
|
--echo * TEST of INSERT vs state.auto_increment
|
WL#3072 - Maria Recovery: recovery of state.auto_increment.
When we log UNDO_KEY_INSERT for an auto_inc key, we update
state.auto_increment (not anymore at the end of maria_write() except
if this is a non-transactional table). When Recovery sees UNDO_KEY_INSERT
in the REDO phase, it reads the auto_inc value from it and updates
state.auto_increment.
mysql-test/r/maria-recovery.result:
Without the code fix, there would be in CHECK TABLE:
"Auto-increment value: 0 is smaller than max used value: 3"
and no AUTO_INCREMENT= clause in SHOW CREATE TABLE.
mysql-test/t/maria-recovery.test:
Test of recovery of state.auto_increment: from an old table,
does the replaying of the log set state.auto_increment to
what it should be.
storage/maria/ma_check.c:
new way of calling ma_retrieve_auto_increment(): pass key
storage/maria/ma_key.c:
ma_retrieve_auto_increment() now operates directly with a pointer
to the key and not on the record.
storage/maria/ma_key_recover.c:
dedicated write_hook_for_undo_key_insert(): sets state.auto_increment
under log's mutex.
storage/maria/ma_key_recover.h:
Dedicated hook for UNDO_KEY_INSERT, to set state.auto_increment.
Such hook needs a new member st_msg_write_hook_for_undo_key::auto_increment,
which contains the auto_increment value inserted.
storage/maria/ma_loghandler.c:
UNDO_KEY_INSERT gets a dedicated write_hook, to set auto_increment.
storage/maria/ma_recovery.c:
When in the REDO phase we see UNDO_KEY_INSERT: if the state is older
than this record, and the key is the auto_increment one, read
the key's value from the log record and update state.auto_increment.
storage/maria/ma_test_all.sh:
use $maria_path to be able to run from /dev/shm (faster)
storage/maria/ma_update.c:
bool is more of C++, using my_bool.
If table is transactional, state.auto_increment is already updated
in write_hook_for_undo_key_insert().
storage/maria/ma_write.c:
If table is transactional, state.auto_increment is not updated at
the end of maria_write() but rather in write_hook_for_undo_key_insert()
(under log's mutex, so that a concurrent checkpoint does not read
state.auto_increment while it is changing - corrupted).
_ma_ck_write_btree_with_log() extracts the auto_increment value
from the key, puts it into msg.auto_increment, and this is passed
to write_hook_for_undo_key_insert().
storage/maria/maria_def.h:
change of prototype of ma_retrieve_auto_increment()
storage/maria/maria_read_log.c:
use default log file size. Use separate page caches for table
and logs (needed if maria_block_size!=TRANSLOG_PAGE_SIZE).
2007-12-12 22:33:36 +01:00
|
|
|
-- source include/maria_empty_logs.inc
|
|
|
|
let $mms_tables=1;
|
|
|
|
CREATE TABLE t1 (
|
|
|
|
i int auto_increment primary key,
|
|
|
|
c varchar(6),
|
|
|
|
key(c)
|
|
|
|
) ENGINE=MARIA;
|
|
|
|
insert into t1 values(null,"b");
|
|
|
|
-- source include/maria_make_snapshot_for_feeding_recovery.inc
|
|
|
|
insert into t1 values(null,"a"), (null,"c"), (null,"d");
|
|
|
|
# With this DELETE we also verify that Recovery cares only about INSERTs
|
|
|
|
delete from t1 where c="d";
|
|
|
|
-- source include/maria_make_snapshot_for_comparison.inc
|
|
|
|
let $mvr_restore_old_snapshot=1;
|
|
|
|
let $mms_compare_physically=0;
|
|
|
|
let $mvr_debug_option="+d,maria_flush_whole_log,maria_crash";
|
|
|
|
let $mvr_crash_statement= set global maria_checkpoint_interval=1;
|
|
|
|
-- source include/maria_verify_recovery.inc
|
|
|
|
show create table t1;
|
|
|
|
|
|
|
|
# Test that UPDATE's effect on auto-increment is recovered
|
WL#3072 - Maria recovery.
* fix for bitmap vs checkpoint bug which could lead to corrupted
tables in case of crashes at certain moments: a bitmap could be flushed
to disk even though it was inconsistent with the log (it could be
flushed before REDO-UNDO are written to the log). One bug remains, need
code from others. Tests added. Fix is to pin unflushable bitmap pages,
and let checkpoint wait for them to be flushable.
* fix for long_trid!=0 assertion failure at Recovery.
* less useless wakeups in the background flush|checkpoint thread.
* store global_trid_generator in checkpoint record.
mysql-test/r/maria-recovery.result:
result update
mysql-test/t/maria-recovery.test:
make it easier to locate subtests
storage/maria/ma_bitmap.c:
When we send a bitmap to the pagecache, if this bitmap is not in a
flushable state we keep it pinned and add it to a list, it will be
unpinned when the bitmap is flushable again.
A new function _ma_bitmap_flush_all() used by checkpoint.
A new function _ma_bitmap_flushable() used by block format to signal
when it starts modifying a bitmap and when it is done with it.
storage/maria/ma_blockrec.c:
When starting a row operation (insert/update/delete), mark that
the bitmap is not flushable (because for example INSERT is going
to over-allocate in the bitmap to prevent other threads from using
our data pages). If a checkpoint comes at this moment it will wait
for the bitmap to be flushable before flushing it.
When the operation ends, bitmap becomes flushable again; that
transition is done under the bitmap's mutex (needed for correct
synchro with a concurrent checkpoint); but for INSERT/UPDATE this
happens inside _ma_bitmap_release_unused() at a place where it already
has the mutex, so the only penalty (mutex adding) is in DELETE and UNDO
of INSERT. In case of errors after setting the bitmap unflushable,
we must always set it back to flushable or checkpoint would block.
Debug possibilities to force a sleep while the bitmap is over-allocated.
In case of error in get_head_or_tail() in allocate_and_write_block_record(),
we still need to unpin all pages.
Bugfix: _ma_apply_redo_insert_row_blobs() produced wrong
data_file_length.
storage/maria/ma_blockrec.h:
new bitmap calls.
storage/maria/ma_checkpoint.c:
filter_flush_indirect not needed anymore (flushing bitmap
pages happens in _ma_bitmap_flush_all() now). So
st_filter_param::is_data_file|pages_covered_by_bitmap not needed.
Other filter_flush* don't need to flush bitmap anymore.
Add debug possibility to flush all bitmap pages outside of a checkpoint,
to simulate pagecache LRU eviction.
When the background flush/checkpoint thread notices it has nothing
to flush, it now sleeps directly until the next potential checkpoint
moment instead of waking up every second.
When in checkpoint we decide to not store a table in the checkpoint record
(because it has logged no writes for example), we can also skip flushing
this table.
storage/maria/ma_commit.c:
comment is out-of-date
storage/maria/ma_key_recover.c:
comment fix
storage/maria/ma_loghandler.c:
comment is out-of-date
storage/maria/ma_open.c:
comment is out-of-date
storage/maria/ma_pagecache.c:
comment for bug to fix. And we don't take checkpoints at end of REDO
phase yet so can trust block->type.
storage/maria/ma_recovery.c:
Comments. Now-unneeded code for incomplete REDO-UNDO groups removed.
When we forget about an old transaction we must really forget
about it with bzero() (fixes the "long_trid!=0 assertion" recovery
bug). When we delete a row with maria_delete() we turn on
STATE_NOT_OPTIMIZED_ROWS so we do the same when we see a CLR_END
for an UNDO_ROW_INSERT or when we execute an UNDO_ROW_INSERT (in both
cases a row was deleted). Pick up max_long_trid from the checkpoint record.
storage/maria/maria_chk.c:
comment
storage/maria/maria_def.h:
MARIA_FILE_BITMAP gets new members: 'flushable', 'bitmap_cond' and
'pinned_pages'.
storage/maria/trnman.c:
I used to think that recovery only needs to know the maximum TrID
of the lists of active and committed transactions. But no, sometimes
both lists can even be empty and their TrID should not be reused.
So Checkpoint now saves global_trid_generator in the checkpoint record.
storage/maria/trnman_public.h:
macros to read/store a TrID
mysql-test/r/maria-recovery-bitmap.result:
result is ok. Without the code fix, we would get a corruption message
about the bitmap page in CHECK TABLE EXTENDED.
mysql-test/t/maria-recovery-bitmap-master.opt:
usual when we crash mysqld in tests
mysql-test/t/maria-recovery-bitmap.test:
test of recovery problems specific of the bitmap pages.
2007-12-14 16:14:12 +01:00
|
|
|
--echo * TEST of UPDATE vs state.auto_increment
|
WL#3072 - Maria Recovery: recovery of state.auto_increment.
When we log UNDO_KEY_INSERT for an auto_inc key, we update
state.auto_increment (not anymore at the end of maria_write() except
if this is a non-transactional table). When Recovery sees UNDO_KEY_INSERT
in the REDO phase, it reads the auto_inc value from it and updates
state.auto_increment.
mysql-test/r/maria-recovery.result:
Without the code fix, there would be in CHECK TABLE:
"Auto-increment value: 0 is smaller than max used value: 3"
and no AUTO_INCREMENT= clause in SHOW CREATE TABLE.
mysql-test/t/maria-recovery.test:
Test of recovery of state.auto_increment: from an old table,
does the replaying of the log set state.auto_increment to
what it should be.
storage/maria/ma_check.c:
new way of calling ma_retrieve_auto_increment(): pass key
storage/maria/ma_key.c:
ma_retrieve_auto_increment() now operates directly with a pointer
to the key and not on the record.
storage/maria/ma_key_recover.c:
dedicated write_hook_for_undo_key_insert(): sets state.auto_increment
under log's mutex.
storage/maria/ma_key_recover.h:
Dedicated hook for UNDO_KEY_INSERT, to set state.auto_increment.
Such hook needs a new member st_msg_write_hook_for_undo_key::auto_increment,
which contains the auto_increment value inserted.
storage/maria/ma_loghandler.c:
UNDO_KEY_INSERT gets a dedicated write_hook, to set auto_increment.
storage/maria/ma_recovery.c:
When in the REDO phase we see UNDO_KEY_INSERT: if the state is older
than this record, and the key is the auto_increment one, read
the key's value from the log record and update state.auto_increment.
storage/maria/ma_test_all.sh:
use $maria_path to be able to run from /dev/shm (faster)
storage/maria/ma_update.c:
bool is more of C++, using my_bool.
If table is transactional, state.auto_increment is already updated
in write_hook_for_undo_key_insert().
storage/maria/ma_write.c:
If table is transactional, state.auto_increment is not updated at
the end of maria_write() but rather in write_hook_for_undo_key_insert()
(under log's mutex, so that a concurrent checkpoint does not read
state.auto_increment while it is changing - corrupted).
_ma_ck_write_btree_with_log() extracts the auto_increment value
from the key, puts it into msg.auto_increment, and this is passed
to write_hook_for_undo_key_insert().
storage/maria/maria_def.h:
change of prototype of ma_retrieve_auto_increment()
storage/maria/maria_read_log.c:
use default log file size. Use separate page caches for table
and logs (needed if maria_block_size!=TRANSLOG_PAGE_SIZE).
2007-12-12 22:33:36 +01:00
|
|
|
-- source include/maria_make_snapshot_for_feeding_recovery.inc
|
|
|
|
update t1 set i=15 where c="a";
|
|
|
|
-- source include/maria_make_snapshot_for_comparison.inc
|
|
|
|
let $mvr_restore_old_snapshot=1;
|
|
|
|
let $mms_compare_physically=0;
|
|
|
|
let $mvr_debug_option="+d,maria_flush_whole_log,maria_crash";
|
|
|
|
let $mvr_crash_statement= set global maria_checkpoint_interval=1;
|
|
|
|
-- source include/maria_verify_recovery.inc
|
|
|
|
show create table t1;
|
Minor changes. New description in SHOW ENGINES for Maria.
Test for BUG#34106 "auto_increment is reset to 1 when table is recovered from crash"
(fixed by Monty yesterday)
mysql-test/r/maria-recovery.result:
result, which is correct (before pulling Monty's fix for BUG#34106,
we got a warning about auto_increment in CHECK TABLE (done in
maria-verify-recovery.inc), no AUTO_INCREMENT clause in
SHOW CREATE TABLE, and a failure of the last INSERT.
mysql-test/r/maria.result:
result
mysql-test/t/maria-recovery.test:
Test for BUG#34106
mysql-test/t/maria.test:
look at what is reported in SHOW ENGINES
mysys/my_pread.c:
changed my mind: if Count argument is >4GB, we'll surely see a segfault
in the pread() call when it tries to read 4GB from memory, so no need
to print it in ulonglong format (saves a function call).
mysys/my_read.c:
changed my mind: if Count argument is >4GB, we'll surely see a segfault
in the pread() call when it tries to read 4GB from memory, so no need
to print it in ulonglong format (saves a function call).
mysys/my_write.c:
changed my mind: if Count argument is >4GB, we'll surely see a segfault
in the pread() call when it tries to read 4GB from memory, so no need
to print it in ulonglong format (saves a function call).
storage/maria/ha_maria.cc:
Description representing the current reality. This can be changed later
storage/maria/ma_page.c:
When reading the new key_del from a page on disk, if there is a bug
(like BUG#34062) this key_del could be wrong, we try to catch if it's
out of the key file.
storage/maria/ma_pagecache.c:
- no truncation of page's number in DBUG_PRINT (useful for BUG#34062)
- page_korr instead of uint5korr
storage/maria/ma_recovery.c:
page_korr instead of uint5korr
storage/maria/plug.in:
Description representing the current reality. This can be changed later.
2008-01-31 23:17:50 +01:00
|
|
|
|
|
|
|
# Test that INSERT's rollback does not set auto-increment counter to 1
|
|
|
|
# (BUG#34106)
|
|
|
|
--echo * TEST of INSERT's rollback vs state.auto_increment
|
|
|
|
-- source include/maria_make_snapshot_for_comparison.inc
|
|
|
|
let $mvr_restore_old_snapshot=0;
|
|
|
|
let $mms_compare_physically=0;
|
|
|
|
let $mvr_debug_option="+d,maria_flush_whole_log,maria_crash";
|
|
|
|
let $mvr_crash_statement= set global maria_checkpoint_interval=1;
|
|
|
|
lock tables t1 write;
|
|
|
|
insert into t1 values(null, "e");
|
|
|
|
-- source include/maria_verify_recovery.inc
|
|
|
|
show create table t1;
|
|
|
|
insert into t1 values(null, "f");
|
WL#3072 - Maria Recovery: recovery of state.auto_increment.
When we log UNDO_KEY_INSERT for an auto_inc key, we update
state.auto_increment (not anymore at the end of maria_write() except
if this is a non-transactional table). When Recovery sees UNDO_KEY_INSERT
in the REDO phase, it reads the auto_inc value from it and updates
state.auto_increment.
mysql-test/r/maria-recovery.result:
Without the code fix, there would be in CHECK TABLE:
"Auto-increment value: 0 is smaller than max used value: 3"
and no AUTO_INCREMENT= clause in SHOW CREATE TABLE.
mysql-test/t/maria-recovery.test:
Test of recovery of state.auto_increment: from an old table,
does the replaying of the log set state.auto_increment to
what it should be.
storage/maria/ma_check.c:
new way of calling ma_retrieve_auto_increment(): pass key
storage/maria/ma_key.c:
ma_retrieve_auto_increment() now operates directly with a pointer
to the key and not on the record.
storage/maria/ma_key_recover.c:
dedicated write_hook_for_undo_key_insert(): sets state.auto_increment
under log's mutex.
storage/maria/ma_key_recover.h:
Dedicated hook for UNDO_KEY_INSERT, to set state.auto_increment.
Such hook needs a new member st_msg_write_hook_for_undo_key::auto_increment,
which contains the auto_increment value inserted.
storage/maria/ma_loghandler.c:
UNDO_KEY_INSERT gets a dedicated write_hook, to set auto_increment.
storage/maria/ma_recovery.c:
When in the REDO phase we see UNDO_KEY_INSERT: if the state is older
than this record, and the key is the auto_increment one, read
the key's value from the log record and update state.auto_increment.
storage/maria/ma_test_all.sh:
use $maria_path to be able to run from /dev/shm (faster)
storage/maria/ma_update.c:
bool is more of C++, using my_bool.
If table is transactional, state.auto_increment is already updated
in write_hook_for_undo_key_insert().
storage/maria/ma_write.c:
If table is transactional, state.auto_increment is not updated at
the end of maria_write() but rather in write_hook_for_undo_key_insert()
(under log's mutex, so that a concurrent checkpoint does not read
state.auto_increment while it is changing - corrupted).
_ma_ck_write_btree_with_log() extracts the auto_increment value
from the key, puts it into msg.auto_increment, and this is passed
to write_hook_for_undo_key_insert().
storage/maria/maria_def.h:
change of prototype of ma_retrieve_auto_increment()
storage/maria/maria_read_log.c:
use default log file size. Use separate page caches for table
and logs (needed if maria_block_size!=TRANSLOG_PAGE_SIZE).
2007-12-12 22:33:36 +01:00
|
|
|
drop table t1;
|
|
|
|
|
* WL#4137 Maria- Framework for testing recovery in mysql-test-run
See test maria-recovery.test for a model; all include scripts have
an "API" section at start if they do take parameters from outside.
* Fixing bug reported by Jani and Monty (when two REDOs about the same
page in one group, see ma_blockrec.c).
* Fixing small bugs in recovery
mysql-test/include/wait_until_connected_again.inc:
be sure to enter the loop (the previous query by the caller may not have
failed: it could be
query;
mysqladmin shutdown;
call this script).
mysql-test/lib/mtr_process.pl:
* Through the "expect" file a test can tell mtr that a server crash
is expected. What the file contains is irrelevant. Now if its last
line starts with "wait", mtr will wait before restarting (it will
wait for the last line to not start with "wait"). This is for
tests which need to mangle files under the feet of a dead mysqld.
* Remove "expect" file before restarting; otherwise there could be a
race condition: tests sees server restarted, does something, writes
an "expect" file, and then mtr removes that file, then
test kills mysqld, and then mtr will never restart it.
storage/maria/ma_blockrec.c:
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- fixing bug in applying of REDO_PURGE_BLOCKS in recovery: page_range
sometimes has TAIL_BIT set, need to turn it down to know the real page
range.
- Both bugs are covered in maria-recovery.test
storage/maria/ma_checkpoint.c:
Capability to, in debug builds only, do some special operations
(flush all bitmap and data pages, flush state, flush log)
and crash mysqld, to later test recovery.
Driven by some --debug=d, symbols.
storage/maria/ma_open.c:
debugging info
storage/maria/ma_pagecache.c:
Now that we can _ma_unpin_all_pages() during the REDO phase
to set page's LSN, the assertion needs to be relaxed.
storage/maria/ma_recovery.c:
- open trace file in append mode (useful when a test triggers several
recoveries, we see them all).
- fixing wrong error detection, it's possible that during recovery
we want to open an already open table.
- when applying a REDO in recovery, we don't anymore put UNDO's LSN on the page
at once; indeed if in this REDO's group there comes another REDO
for the same page it would be wrongly skipped. Instead, we keep
pages pinned, don't change their LSN. When done with all REDOs
of the group we unpin them and stamp them with UNDO's LSN.
- we verify that all log records of a group are about the same table,
for debugging.
mysql-test/r/maria-recovery.result:
result
mysql-test/t/maria-recovery-master.opt:
crash is expected, core file would take room, stack trace would
wake pushbuild up.
mysql-test/t/maria-recovery.test:
Test of recovery from mysql-test (it is already tested as unit tests
in ma_test_recovery) (WL#4137)
- test that, if recovery is made to start on an empty table it can
replay the effects of committed and uncommitted statements (having only
the committed ones in the end result). This should be the first test
for someone writing code of new REDOs.
- test that, if mysqld is crashed and recovery runs we have only
committed statements in the end result. Crashes are done in different
ways: flush nothing (so, uncommitted statement is often missing
from the log => no rollback to do); flush pagecache (implicitely flushes
log (WAL)) and flush log, both causes rollbacks; flush log can also
flush state (state.records etc) to test recovery of the state
(not tested well now as we repair the index anyway).
- test of bug found by Jani and Monty in recovery (two REDO about
the same page in one group).
mysql-test/include/maria_empty_logs.inc:
removes logs, to have a clean sheet for testing recovery.
mysql-test/include/maria_make_snapshot.inc:
copies a table to another directory, or back, or compares both
(comparison is not implemented as physical comparison is impossible
if an UNDO phase happened).
mysql-test/include/maria_make_snapshot_for_comparison.inc:
copies tables to another directory so that they can later
serve as a comparison reference (they are the good tables,
recovery should produce similar ones).
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
When we want to force recovery to start on old tables, we prepare
old tables with this script: we put them in a spare directory.
They are later copied back over mysqltest tables while mysqld is dead.
We also need to copy back the control file, otherwise mysqld,
in recovery, would start from the latest checkpoint: latest
checkpoint plus old tables is not a recovery-possible scenario of course.
mysql-test/include/maria_verify_recovery.inc:
causes mysqld to crash, restores old tables if requested,
lets recovery run, compares resulting tables with reference tables
by using CHECKSUM TABLE.
We don't do any sanity checks on page's LSN in resulting tables, yet.
2007-11-13 17:12:29 +01:00
|
|
|
# clean up everything
|
|
|
|
let $mms_purpose=feeding_recovery;
|
|
|
|
eval drop database mysqltest_for_$mms_purpose;
|
|
|
|
let $mms_purpose=comparison;
|
|
|
|
eval drop database mysqltest_for_$mms_purpose;
|
|
|
|
drop database mysqltest;
|