mirror of
https://github.com/MariaDB/server.git
synced 2025-01-31 02:51:44 +01:00
19 commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
unknown
|
3563311bfd |
Merge gbichot@bk-internal.mysql.com:/home/bk/mysql-maria
into gbichot4.local:/home/mysql_src/mysql-maria-monty mysql-test/t/maria-recovery.test: Auto merged storage/maria/ha_maria.cc: Auto merged storage/maria/ma_loghandler.c: Auto merged storage/maria/ma_pagecache.c: Auto merged |
||
|
unknown
|
2fcff8988a |
Fix for BUG#34114 "maria_chk reports false error when several tables on
command-line" and BUG#34062 "Maria table corruption on master". Use 5 bytes (instead of 4) to store page's number in the checkpoint record, to allow bigger table (1PB with maria-block-size=1kB). Help pushbuild not run out of memory by moving the portion of maria-recovery.test which generates lots of data into a -big.test. mysql-test/r/maria-recovery.result: result moved mysql-test/t/maria-recovery.test: piece which generates much data moved to maria-recovery-big.test mysys/my_pread.c: To fix BUG#34062, where a 1.1TB file was generated due to a wrong pwrite offset, it was useful to not lose precision on 'offset' in DBUG_PRINT, so that the crazy value is visible. mysys/my_read.c: To fix BUG#34062, where a 1.1TB file was generated due to a wrong pwrite offset, it was useful to not lose precision on 'offset' in DBUG_PRINT, so that the crazy value is visible. mysys/my_write.c: To fix BUG#34062, where a 1.1TB file was generated due to a wrong pwrite offset, it was useful to not lose precision on 'offset' in DBUG_PRINT, so that the crazy value is visible. storage/maria/ha_maria.cc: When starting a bulk insert, we throw away dirty index pages from the cache. Unique (non disabled) key insertions thus read out-of-date pages from the disk leading to BUG#34062 "Maria table corruption on master": a DELETE in procedure viewer_sp() had deleted all rows of viewer_tbl2 one by one, putting index page 1 into key_del; that page was thrown away at start of INSERT SELECT, then the INSERT SELECT needed a page to insert keys, looked at key_del, found 1, read page 1 from disk, and its out-of-date content was used to set the new value of key_del (crazy value of 1TB), then a later insertion needed another index page, tried to read page at this crazy offset and failed, leading to corruption mark. The fix is to destroy out-of-date pages and make the state consistent with that, i.e. call maria_delete_all_rows(). storage/maria/ma_blockrec.c: Special hook for UNDO_BULK_INSERT storage/maria/ma_blockrec.h: special hook for UNDO_BULK_INSERT storage/maria/ma_check.c: Fix for BUG#34114 "maria_chk reports false error when several tables on command-line": if the Nth (on the command line) table was BLOCK_RECORD it would start checks by using the param->record_checksum computed by checks of table N-1. storage/maria/ma_delete_all.c: comment storage/maria/ma_loghandler.c: special hook for UNDO_BULK_INSERT storage/maria/ma_page.c: comment storage/maria/ma_pagecache.c: page number is 5 bytes in checkpoint record now (allows bigger tables) storage/maria/ma_recovery.c: page number is 5 bytes in checkpoint record now storage/maria/ma_recovery_util.c: page number is 5 bytes now storage/maria/ma_write.c: typo mysql-test/r/maria-recovery-big.result: result is correct mysql-test/t/maria-recovery-big-master.opt: usual options for recovery tests mysql-test/t/maria-recovery-big.test: Moving out the big blob test to a -big test (it exhausts memory when using /dev/shm on certain machines) |
||
|
unknown
|
a1255fc299 |
Support of moving logs in other directory then data root.
Fixed typo in maria_read_log "help" output.
Compilation warning on Mac OS fixed.
mysql-test/include/maria_empty_logs.inc:
Added support of moving maria log relatively
to master data directory.
mysql-test/include/maria_make_snapshot_for_feeding_recovery.inc:
Added support of moving maria log relatively
to master data directory.
mysql-test/include/maria_verify_recovery.inc:
Added support of moving maria log relatively
to master data directory.
mysql-test/t/maria-recovery-master.opt:
Check of recovery with logs in other place.
mysql-test/t/maria-recovery.test:
Check of recovery with logs in other place.
storage/maria/ha_maria.cc:
Support of moving logs in other directory then data root.
storage/maria/ma_loghandler.c:
Compilation warning on Mac OS fixed.
storage/maria/ma_static.c:
Support of moving logs in other directory then data root.
storage/maria/ma_test1.c:
Support of moving logs in other directory then data root.
storage/maria/maria_def.h:
Support of moving logs in other directory then data root.
storage/maria/maria_read_log.c:
Support of moving logs in other directory then data root.
Fixed typo in maria_read_log "help" output.
|
||
|
unknown
|
534417c285 |
WL#3072 - Maria Recovery
Bulk insert: don't log REDO/UNDO for rows, log one UNDO which will truncate files; this is an optimization and a bugfix (table was left half-repaired by crash). Repair: mark table crashed-on-repair at start, bump skip_redo_lsn at start, this is easier for recovery (tells it to skip old REDOs or even UNDO phase) and user (tells it to repair) in case of crash, sync files in the end. Recovery skips missing or corrupted table and moves to next record (in REDO or UNDO phase) to be more robust; warns if happens in UNDO phase. Bugfix for UNDO_KEY_DELETE_WITH_ROOT (tested in ma_test_recovery) and maria_enable_indexes(). Create missing bitmaps when needed (there can be more than one to create, in rare cases), log a record for this. include/myisamchk.h: new flag: bulk insert repair mustn't bump create_rename_lsn mysql-test/lib/mtr_report.pl: skip normal warning in maria-recovery.test mysql-test/r/maria-recovery.result: result: crash before bulk insert is committed, causes proper rollback, and crash right after OPTIMIZE replaces index file with new index file leads to table marked corrupted and recovery not failing. mysql-test/t/maria-recovery.test: - can't check the table or it would commit the transaction, but check is made after recovery. - test of crash before bulk-insert-with-repair is committed (to see if it is rolled back), and of crash after OPTIMIZE has replaced index file but not finished all operations (to see if recovery fails - it used to assert when trying to execute an old REDO on the new index). storage/maria/CMakeLists.txt: new file storage/maria/Makefile.am: new file storage/maria/ha_maria.cc: - If bulk insert on a transactional table using an index repair: table is initially empty, so don't log REDO/UNDO for data rows (optimization), just log an UNDO_BULK_INSERT_WITH_REPAIR which will, if executed, empty the data and index file. Re-enable logging in end_bulk_insert(). - write log record for repair operation only after it's fully done, index sort including (maria_repair*() used to write the log record). - Adding back file->trn=NULL which was removed by mistake earlier. storage/maria/ha_maria.h: new member (see ha_maria.cc) storage/maria/ma_bitmap.c: Functions to create missing bitmaps: - one function which creates missing bitmaps in page cache, except the missing one with max offset which it does not put into page cache as it will be modified very soon. - one function which the one above calls, and creates bitmaps in page cache - one function to execute REDO_BITMAP_NEW_PAGE which uses the second one above. storage/maria/ma_blockrec.c: - when logging REDO_DELETE_ALL, not only 'records' and 'checksum' has to be reset under log's mutex. - execution of REDO_INSERT_ROW_BLOBS now checks the dirty pages' list - execution of UNDO_BULK_INSERT_WITH_REPAIR storage/maria/ma_blockrec.h: new functions storage/maria/ma_check.c: - table-flush-before-repair is moved to a separate function reused by maria_sort_index(); syncing is added - maria_repair() is allowed to re-enable logging only if it is the one which disabled it. - "_ma_flush_table_files_after_repair" was a bad name, it's not after repair now, and it should not sync as we do more changes to the files shortly after (sync is postponed to when writing the log record) - REDO_REPAIR record should be written only after all repair operations (in particular after sorting index in ha_mara::repair()) - close to the end of repair by sort, flushing of pages must happen also in the non-quick case, to prepare for the sync at end. - in parallel repair, some page flushes are not needed as done by initialize_variables_for_repair(). storage/maria/ma_create.c: Update skip_redo_lsn, create_rename_lsn optionally. storage/maria/ma_delete_all.c: Need to sync files at end of maria_delete_all_rows(), if transactional. storage/maria/ma_extra.c: During repair, we sometimes call _ma_flush_table_files() (via _ma_flush_table_files_before_swap()) while there is a WRITE_CACHE. storage/maria/ma_key_recover.c: - when we see CLR_END for UNDO_BULK_INSERT_WITH_REPAIR, re-enable indices. - fixing bug: _ma_apply_undo_key_delete() parsed UNDO_KEY_DELETE_WITH_ROOT wrongly, leading to recovery failure storage/maria/ma_key_recover.h: new prototype storage/maria/ma_locking.c: DBUG_VOID_RETURN missing storage/maria/ma_loghandler.c: UNDO for bulk insert with repair, and REDO for creating bitmaps. LOGREC_FIRST_FREE to not have to change the for() every time we add a new record type. storage/maria/ma_loghandler.h: new UNDO and REDO storage/maria/ma_open.c: Move share.kfile.file=kfile up a bit, so that _ma_update_state_lsns() can get its value, this fixes a bug where LSN_REPAIRED_BY_MARIA_CHK was not corrected on disk by maria_open(). Store skip_redo_lsn in index' header. maria_enable_indexes() had a bug for BLOCK_RECORD, where an empty file has one page, not 0 bytes. storage/maria/ma_recovery.c: - Skip a corrupted, missing, or repaired-with-maria_chk, table in recovery: don't fail, just go to next REDO or UNDO; but if an UNDO is skipped in UNDO phase we issue warnings. - Skip REDO|UNDO in REDO phase if <skip_redo_lsn. - If UNDO phase fails, delete transactions to not make trnman assert. - Update skip_redo_lsn when playing REDO_CREATE_TABLE - Don't record UNDOs for old transactions which we don't know (long_trid==0) - Bugfix for UNDO_KEY_DELETE_WITH_ROOT (see ma_key_recover.c) - Execution of UNDO_BULK_INSERT_WITH_REPAIR - Don't try to find a page number in REDO_DELETE_ALL - Pieces moved to ma_recovery_util.c storage/maria/ma_rename.c: name change storage/maria/ma_static.c: I modified layout of the index' header (inserted skip_redo_lsn in its middle) storage/maria/ma_test2.c: allow breaking the test towards the end, tests execution of UNDO_KEY_DELETE_WITH_ROOT storage/maria/ma_test_recovery.expected: 6 as testflag instead of 4 storage/maria/ma_test_recovery: Increase the amount of rollback work to do when testing recovery with ma_test2; this reproduces the UNDO_KEY_DELETE_WITH_ROOT bug. storage/maria/maria_chk.c: skip_redo_lsn should be updated too, for consistency. Write a REDO_REPAIR after all operations (including sort-records) have been done. No reason to flush blocks after maria_chk_data_link() and maria_sort_records(), there is maria_close() in the end. write_log_record() is a function, to not clutter maria_chk(). storage/maria/maria_def.h: New member skip_redo_lsn in the state, and comments storage/maria/maria_pack.c: skip_redo_lsn should be updated too, for consistency storage/maria/ma_recovery_util.c: _ma_redo_not_needed_for_page(), defined in ma_recovery.c, is needed by ma_blockrec.c; this causes link issues, resolved by putting _ma_redo_not_needed_for_page() into a new file (so that it is not in the same file as repair-related objects of ma_recovery.c). storage/maria/ma_recovery_util.h: new file |
||
|
unknown
|
17f0738885 |
WL#3072 Maria Recovery
All statements doing an implicit commit now also do one in Maria. This is useful because LOCK TABLES; REPAIR; crash; is not rollback-able, the implicit commit of REPAIR avoid that Recovery tries to rollback and fails. Fix for BUG#33827 "COMMIT AND CHAIN causes serious Valgrind error" (maybe not the definite one, depends on the assigned dev). mysql-test/t/maria-recovery.test: test of REPAIR's implicit commit. I cannot commit the result file because maria-recovery fails in vanilla tree (seen in pushbuild) but its new section looks like: repair table t1; Table Op Msg_type Msg_text mysqltest.t1 repair status OK insert into t1 values(2); select * from t1; a 1 2 3 SET SESSION debug="+d,maria_flush_whole_log,maria_flush_whole_page_cache,maria_crash"; * crashing mysqld intentionally set global maria_checkpoint_interval=1; ERROR HY000: Lost connection to MySQL server during query * recovery happens check table t1 extended; Table Op Msg_type Msg_text mysqltest.t1 check status OK * testing that checksum after recovery is as expected Checksum-check failure use mysqltest; select * from t1; a 1 3 Which is as it should be. sql/rpl_injector.cc: fix for BUG#33827 sql/sql_parse.cc: - All DDLs and mysql_admin_table() (REPAIR etc) use end_actrive_trans() to do an implicit commit so we add there an implicit commit of the Maria transaction. - Fix for BUG#33827 storage/maria/ha_maria.cc: - A method to do implicit commit in Maria - After an implicit commit, if it was under LOCK TABLES, the locked tables have a stale file->trn: update it. storage/maria/ha_maria.h: new static method storage/maria/ma_check.c: bugfix: this disabling of transactionality had the effect that if LOCK TABLES; REPAIR; INSERT then the INSERT ran non-transactional (so couldn't be undone in case of crash, if, by bad chance, its effect on pages went to disk). storage/maria/ma_checkpoint.c: indentation storage/maria/ma_recovery.c: dbug statements storage/maria/trnman.c: When doing an implicit commit we need to know the number of locked tables of the committed transaction and copy it to the new transaction storage/maria/trnman_public.h: prototype change |
||
|
unknown
|
266fde77b2 |
Added --loose-skip-maria to MYSQLD_BOOTSTRAP_CMD to get bootstrap.test to work
Allow one to run bootstrap even if --skip-maria is used (needed for bootstrap.test) Fixed lots of compiler warnings NOTE: maria-big and maria-recover tests failes becasue of bugs in transaction log handling. Sanja knows about this and is working on it! mysql-test/mysql-test-run.pl: Added --loose-skip-maria to MYSQLD_BOOTSTRAP_CMD to get bootstrap.test to work mysql-test/r/maria-recovery.result: Updated results mysql-test/t/bootstrap.test: Removed not needed empty line mysql-test/t/change_user.test: Fixed results for 32 bit systems mysql-test/t/maria-big.test: Only run this when you use --big mysql-test/t/maria-recovery.test: Added test case for recovery with big blobs mysys/my_uuid.c: Fixed compiler warning sql/mysqld.cc: Allow one to run bootstrap even if --skip-maria is used (needed for bootstrap.test) sql/set_var.cc: Compare max_join_size with ULONG_MAX instead of HA_POS_ERROR as we set max_join_size to ULONG_MAX by default storage/maria/ma_bitmap.c: Added __attribute((unused)) to fix compiler warning storage/maria/ma_blockrec.c: Added casts to remove compiler warnings Change variable types to avoid compiler warnings storage/maria/ma_check.c: Added casts to remove compiler warnings storage/maria/ma_checkpoint.c: Change variable types to avoid compiler warnings storage/maria/ma_create.c: Change variable types to avoid compiler warnings storage/maria/ma_delete.c: Added casts to remove compiler warnings storage/maria/ma_key_recover.c: Added casts to remove compiler warnings storage/maria/ma_loghandler.c: Moved initiazation of prev_buffer first as this could otherwise not be set in case of errors storage/maria/ma_page.c: Added casts to remove compiler warnings storage/maria/ma_pagecache.c: Added __attribute((unused)) to fix compiler warning storage/maria/ma_pagecrc.c: Added #ifndef DBUG_OFF to remove compiler warning storage/maria/ma_recovery.c: Added casts to remove compiler warnings storage/maria/ma_write.c: Added casts to remove compiler warnings storage/maria/maria_chk.c: Split long string into two to avoid compiler warnings storage/myisam/ft_boolean_search.c: Added LINT_INIT() to remove compiler warning support-files/compiler_warnings.supp: Suppress wrong compiler warning unittest/mytap/tap.c: Fixed declaration to match prototypes to remove compiler warnings |
||
|
unknown
|
98aad88f4d |
after-merge fixes and comments
mysql-test/include/maria_empty_logs.inc: At one moment in maria-recovery.test the first log has number 2, because log 1 was manually deleted. mysql-test/r/maria-recovery.result: after-merge fix mysql-test/t/maria-recovery.test: after-merge fix storage/maria/ma_bitmap.c: after-merge fix. The todo is implemented now. storage/maria/ma_blockrec.c: comment storage/maria/ma_open.c: after-merge fix. Set write_fail also for index file or a write error would crash. storage/maria/ma_pagecache.c: comment storage/maria/ma_pagecache.h: I prefer to use NULL for 'no callback' instead of a dummy callback in the special case of get_log_address; indeed for non-transactional tables it uses an if(), while if using a dummy callback, it would use a function call plus an if() (the dummy callback would need to return a magic value to say "don't flush" and that value would be tested in if()). storage/maria/unittest/ma_test_all-t: fix if running from outside storage/maria |
||
|
unknown
|
75a8907ee0 |
Merge gbichot@bk-internal.mysql.com:/home/bk/mysql-maria
into gbichot4.local:/home/mysql_src/mysql-maria-for-undo-phase storage/maria/ma_loghandler.c: Auto merged storage/maria/ma_recovery.c: Auto merged mysql-test/r/maria-recovery.result: merge mysql-test/t/maria-recovery.test: merge |
||
|
unknown
|
e580e058e5 |
Maria: more log handler fixes for when logs manually deleted
mysql-test/include/maria_empty_logs.inc: can now ask that control file be preserved mysql-test/lib/mtr_report.pl: expected warning (produced by translog_init() when log is missing) mysql-test/r/maria-recovery.result: result update mysql-test/t/maria-recovery.test: test of removing logs manually; recovery used to fail storage/maria/ma_loghandler.c: The addition to maria-recovery.test uncovered a bug: log 1 was manually deleted, Maria restarted, created log 2, then crash before a checkpoint: logs_found is TRUE, last_checkpoint_lsn is LSN_IMPOSSIBLE, so sure_page assumes that log 1 exists; using last_logno instead (only the last log needs to be checked, as the 'else' branch checks only after last_checkpoint_lsn: creation of a log is as good a starting point as a checkpoint, for scanning the log). |
||
|
unknown
|
1e9ee8abbd |
Merge bk-internal.mysql.com:/home/bk/mysql-maria
into mysql.com:/home/my/mysql-maria dbug/dbug.c: Auto merged include/my_dbug.h: Auto merged include/my_global.h: Auto merged mysql-test/mysql-test-run.pl: Auto merged mysql-test/r/maria-recovery.result: Auto merged mysql-test/t/maria-recovery.test: Auto merged mysql-test/t/maria.test: Auto merged sql/mysqld.cc: Auto merged sql/sql_show.cc: Auto merged storage/maria/Makefile.am: Auto merged storage/maria/ha_maria.cc: Auto merged storage/maria/ma_bitmap.c: Auto merged storage/maria/ma_check.c: Auto merged storage/maria/ma_key_recover.c: Auto merged storage/maria/ma_loghandler.c: Auto merged storage/maria/ma_open.c: Auto merged storage/maria/ma_recovery.c: Auto merged storage/maria/maria_chk.c: Auto merged storage/maria/maria_def.h: Auto merged storage/maria/ma_test_all.sh: Use remote version mysql-test/r/maria.result: Trivial merge storage/maria/ma_pagecache.c: Trivial merge |
||
|
unknown
|
bfd5c273c4 |
UNDO of rows now puts back all part of the row on their original pages and positions
Added variable _dbug_on_ to speed up execution when DBUG is not going to be used Added --debug-on option to mysqld (to be able to turn of DBUG with --debug-on=0) Fixed some bugs with 'non_flushable' marking of bitmap pages Don't use 'non_flushable' marking of bitmap pages for not transactional tables SHOW CREATE TABLE now shows if table was created with page checksums Fixed a lot of bugs with BLOB handling in case of update/REDO and UNDO More tests (especially for blobs) and DBUG_ASSERTS() More readable output from maria_read_log and maria_chk Fixed wrong shift that caused Maria to crash on files > 4G Mark tables as crashed of REDO fails dbug/dbug.c: Changed to use my_bool (allowed me to remove some windows specific code) Added variable _dbug_on_ to speed up execution when DBUG is not going to be used Removed initialization of variables if not needed include/my_dbug.h: Use my_bool for some functions that was defined as BOOLEAN in dbug.c code Added DBUGGER_ON/DEBUGGER_OFF to speed up execution when DBUG is not used include/my_global.h: Define my_bool early Increase MY_HOW_OFTEN_TO_WRITE as computers are now faster than 10 years ago mysql-test/mysql-test-run.pl: Added debug-on=0 to speed up tests mysql-test/r/maria-recovery.result: Added new test by Guilhem to test if UNDO_ROW_DELETE preserves rowid mysql-test/r/maria.result: Added testing of page checksums mysql-test/t/crash_commit_before-master.opt: Added --debug-on as test require DBUG to work mysql-test/t/maria-recovery-bitmap-master.opt: Added --debug-on as test require DBUG to work mysql-test/t/maria-recovery-master.opt: Added --debug-on as test require DBUG to work mysql-test/t/maria-recovery.test: Added new test by Guilhem to test if UNDO_ROW_DELETE preserves rowid mysql-test/t/maria.test: Added testing of page checksums sql/mysqld.cc: Added --debug-on option (to be able to turn of DBUG with --debug-on=0) Indentation fixes Removed end spaces sql/sql_show.cc: Allow update_create_info() to inform MySQL if PACK_KEYS, NO_PACK_KEYS, CHECKSUM, PAGE_CHECKSUM or DELAY_KEY_WRITE is used storage/maria/Makefile.am: Added ma_test_big.sh storage/maria/ha_maria.cc: Store in create_info if page checksums are used (For SHOW CREATE TABLE) storage/maria/ma_bitmap.c: Added _ma_bitmap_wait_or_flush() to cause reader of bitmap pages to wait with reading until bitmap is flushed. Use TAIL_PAGE_COUNT_MARKER for tail pages Set 'sub_blocks' for and only for the head page or for the first extent of a blob. This is needed for store_extent_info() to be able to set START_EXTENT_BIT's Don't allocate more than 0x3ffff pages in one extent (We need bit 0x4000 as a START_EXTENT_BIT) Increase the calculated 'head_length' with the number of bytes used for extents. Update row->space_on_head_page also in _ma_bitmap_find_new_place() Make _ma_bitmap_get_page_bits() global. (Needed for UNDO handling) Changed _ma_bitmap_flushable() to take MARIA_HA instead of MARIA_SHARE. This was needed to be able to mark the handler if we had a 'non_flushable' call pending or not. Don't use 'non_flushable' marking of bitmap pages for not transactional tables. Added BLOCKUSED_USE_ORG_BITMAP handling also for tail pages. Added more DBUG_ASSERT() to find possible errors in other code Some code simplications by adding new local variables storage/maria/ma_blockrec.c: UNDO of rows now puts back all part of the row on their original pages and positions. Changed UNDO of DELETE and UNDO of UPDATE to contain information about the original length of data on head block and also extent information This changes a lot of logic as now an insert of a row on a page may happen to any position (and not just to the first or next free) Use PAGE_COUNT to mark if an extent is the start of of a blob. (Needed for extent_to_bitmap_blocks()) Added check_directory() for checking that directroy entries are correct. Added checking of row checksums when reading rows (with EXTRA_DEBUG) Added make_space_for_directory() and extend_directory() for doing expansion of directory Added get_rowpos_in_head_or_tail_page() to be able to store head/tail on original position in UNDO Added extent_to_bitmap_blocks() to be able to generate original bitmap blocks from UNDO entry Added _ma_update_at_original_place() for UNDO of DELETES Added row->min_length to hold minmum required space needed on head page Changed find_free_position() to use make_space_for_directory() Changed make_empty_page() to allow optional creation of directory entry Changed delete_head_or_tail() and _ma_apply_undo_row_isnert() to not copy pagecache block (speed optimization) Changed _ma_apply_redo_insert_row_head_or_tail() to be able to insert new row at any position on 'new' page Changed _ma_apply_undo_row_delete() and _ma_apply_undo_row_update() to put row in it's original position Ensure allocation of tail blocks are of at least MIN_TAIL_SIZE. Ensure we store pages in pinned pages even if read failed. (If not we will have pages pinned forever in page cache) Write original extent information in UNDO entry, not compacted ones (we need position to tails!) When setting BLOCKUSED_USED, don't clear other bits (we have to preserve BLOCKUSED_USE_ORG_BITMAP) Fixed som bugs in directory handling Fixed bug where we wrote wrong lsn to blob pages Added separate blob_buffer for fixing bug when updating row that had char/varchar that spanned several pages and also had blobs Ensure we call _ma_bitmap_flushable() also in case of errors When doing an update, first delete old entries, then search in bitmap for where to put new information Info->s -> share Rowid -> rowid More DBUG_ASSERT() storage/maria/ma_blockrec.h: Added START_EXTENT_BIT and TAIL_PAGE_COUNT_MARKER Added _ma_bitmap_wait_or_flush() and _ma_bitmap_get_page_bits() storage/maria/ma_check.c: Don't write extra empty line if there is no deleted blocks Ignore START_EXTENT_BIT's in page count Call _ma_fast_unlock_key_del() to free key_del link storage/maria/ma_close.c: Ensure that used_key_del is 0. (If not, someone forgot to call _ma_unlock_key_del()) storage/maria/ma_create.c: Changed constant to macro storage/maria/ma_delete.c: For deleted keys, log also position to row storage/maria/ma_extra.c: Release blob buffer at maria_reset() if bigger than MARIA_SMALL_BLOB_BUFFER storage/maria/ma_key_recover.c: Added bzero() of LSN that confused paged cache in case of uninitialized block Mark file crashed if applying of index changes fails Added calls to _ma_fast_unlock_key_del() for protection of shared key_del link. storage/maria/ma_locking.c: Added usage of MARIA_FILE_OPEN_COUNT_OFFSET Added _ma_mark_file_crashed() storage/maria/ma_loghandler.c: Fixed bug where we logged uninitialized memory storage/maria/ma_open.c: Moved state->changed to be at start of state info on disk to allow one to easly mark files as crashed storage/maria/ma_page.c: Disable 'dummy' checksumming of pages as this gave false warnings. (Need to investigate if this is ever needed) storage/maria/ma_pagecache.c: Fixed wrong shift that caused Maria to crash on files > 4G storage/maria/ma_recovery.c: In case of errors, start writing on new line if we where in %## %## printing mode (Made errors more readable) Changed global variable name from warnings -> recovery_warnings Use MARIA_FILE_CREATE_RENAME_LSN_OFFSET instead of constant Removed special handling of row position for deleted keys. Keys now always includes row positions _ma_apply_undo_row_delete() now gets page and row position Added check that we don't loop forever when handling undo's (in case of bug in undo chain) Print name of failed REDO/UNDO storage/maria/ma_recovery.h: Removed old comment storage/maria/ma_static.c: Chaned version number of Maria files to not accidently use old ones (becasue of change of ordering of status variables) storage/maria/ma_test2.c: Added option -u to specify number of rows to update Changed old option -u to be -A, as for ma_test1 Fixed bug in update of rows with blobs (before blobs was always reset to empty on update) First created blob is now of max blob length to ensure we have at least one big blob in the table storage/maria/ma_test_all.sh: More tests storage/maria/ma_test_recovery.expected: Updated results storage/maria/ma_test_recovery: Changed tests to use bigger blobs (not just 1K) Added new tests that tests recovery of update with blobs Removed comparision of .MAD file as it's not guranteed that recovery from scratch gives identical data file as original update (compact_page() may be called at different times during normal execution and during REDO) storage/maria/ma_update.c: Simplify code (changed * to if) storage/maria/maria_chk.c: Make output more readable storage/maria/maria_def.h: Changed 'changed' to int to prepare for more bits Added 2 more bytes to status information Added 'st_mara_row->min_length' for storing min length needed on head page Added 'st_mara_handler->blob_buff & blob_buff_size' for storing blobs Moved all tunning parameters into one block Added MARIA_SMALL_BLOB_BUFFER Added _ma_mark_file_crashed() storage/myisam/mi_test2.c: Fixed bug in update of rows with blobs (before blobs was always reset to empty on update) storage/maria/ma_test_big.sh: Testing of insert, update, delete, recovery and undo of rows with blobs Thanks to the random-ness of ma_test2 this is likely to find most bugs in the row handling |
||
|
unknown
|
389dcccbed |
- Clean up function for my_uuid()
- Merge fixes - "make test" for maria - Replaced ma_test_all with a perl version, which now can be run in unittest mode. include/my_sys.h: Added clean up function for my_uuid() mysql-test/r/create.result: Fixed result file. mysql-test/r/maria-big.result: Changed not to get a warning. mysql-test/r/maria-connect.result: Not to get a warning. mysql-test/r/maria-recovery.result: Not to get a warning. mysql-test/r/maria.result: Fixed result file. Added test for warning on log file. mysql-test/r/ps_maria.result: Fixed result file. New bit was added to source. mysql-test/t/maria-big.test: To avoid a warning. mysql-test/t/maria-connect.test: To avoid a warning. mysql-test/t/maria-recovery.test: To avoid a warning. mysql-test/t/maria.test: To avoid a warning. mysql-test/t/ps_maria.test: To avoid a warning. mysys/my_uuid.c: Added clean up function for my_uuid() sql/mysqld.cc: Manual merge. storage/maria/Makefile.am: Added "make test" for maria. storage/maria/ma_key_recover.c: Fix for maria_recovery storage/maria/ma_test_all.sh: Deprecated file. Functionality moved to unittest/ma_test_all-t BitKeeper/deleted/.del-unit.pl: BitKeeper file /home/my/mysql-maria/storage/maria/unit.pl |
||
|
unknown
|
d72c22dee4 |
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. |
||
|
unknown
|
6b89901cdf |
Merge desktop.sanja.is.com.ua:/home/bell/mysql/bk/mysql-maria
into desktop.sanja.is.com.ua:/home/bell/mysql/bk/work-maria-testfail mysql-test/t/maria-recovery.test: Auto merged storage/maria/ma_loghandler.c: Auto merged |
||
|
unknown
|
ebafe3e4db |
Postreview changes.
mysql-test/r/maria-big.result:
Change to make maria-purge more stable: increasing file size
to let all records fit in one log.
mysql-test/r/maria-connect.result:
Change to make maria-purge more stable: increasing file size
to let all records fit in one log.
mysql-test/r/maria-purge.result:
Change to make maria-purge more stable.
mysql-test/r/maria.result:
Change to make maria-purge more stable: increasing file size
to let all records fit in one log.
mysql-test/r/ps_maria.result:
Change to make maria-purge more stable: increasing file size
to let all records fit in one log.
mysql-test/t/maria-big.test:
Change to make maria-purge more stable: increasing file size
to let all records fit in one log.
mysql-test/t/maria-connect.test:
Change to make maria-purge more stable: increasing file size
to let all records fit in one log.
mysql-test/t/maria-purge.test:
Change to make maria-purge more stable.
mysql-test/t/maria-recovery.test:
Change to make maria-purge more stable: increasing file size
to let all records fit in one log.
mysql-test/t/maria.test:
Change to make maria-purge more stable: increasing file size
to let all records fit in one log.
mysql-test/t/ps_maria.test:
Change to make maria-purge more stable: increasing file size
to let all records fit in one log.
|
||
|
unknown
|
f7df9addba |
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). |
||
|
unknown
|
13f45b160b |
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). |
||
|
unknown
|
ac590d9ce9 |
Addendum to WL#4137 "Maria- Framework for testing recovery in mysql-test-run":
possibility, if requested in the test, to do byte-for-byte comparison of data files after recovery. This makes sense only when testing REDOs, as UNDO phase generates new log records thus data pages get new LSNs. First test of maria-recovery.test is changed to test only REDOs and do byte-for-byte comparison. mysql-test/include/maria_make_snapshot.inc: possibility to compare physically the data files (makes sense only if testing REDO phase). mysql-test/include/maria_verify_recovery.inc: allow physical comparison of data files after recovery if requested mysql-test/r/maria-recovery.result: result update (first test tests REDO only now) mysql-test/t/maria-recovery.test: first test tests REDO only now, and thus can do byte-for-byte comparison of data files. |
||
|
unknown
|
5fbd5dafe7 |
* 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. |