NOTE: Backporting the patch to next-mr.
The fix proposed in BUG#35542 and BUG#31665 introduces a performance issue
when fsyncing the master.info, relay.info and relay-log.bin* after #th events.
Although such solution has been proposed to reduce the probability of corrupted
files due to a slave-crash, the performance penalty introduced by it has
made the approach impractical for highly intensive workloads.
In a nutshell, the option --syn-relay-log proposed in BUG#35542 and BUG#31665
simultaneously fsyncs master.info, relay-log.info and relay-log.bin* and
this is the main source of performance issues.
This patch introduces new options that give more control to the user on
what should be fsynced and how often:
1) (--sync-master-info, integer) which syncs the master.info after #th event;
2) (--sync-relay-log, integer) which syncs the relay-log.bin* after #th
events.
3) (--sync-relay-log-info, integer) which syncs the relay.info after #th
transactions.
To provide both performance and increased reliability, we recommend the following
setup:
1) --sync-master-info = 0 eventually the operating system will fsync it;
2) --sync-relay-log = 0 eventually the operating system will fsync it;
3) --sync-relay-log-info = 1 fsyncs it after every transaction;
Notice, that the previous setup does not reduce the probability of
corrupted master.info and relay-log.bin*. To overcome the issue, this patch also
introduces a recovery mechanism that right after restart throws away relay-log.bin*
retrieved from a master and updates the master.info based on the relay.info:
4) (--relay-log-recovery, boolean) which enables a recovery mechanism that
throws away relay-log.bin* after a crash.
However, it can only recover the incorrect binlog file and position in master.info,
if other informations (host, port password, etc) are corrupted or incorrect,
then this recovery mechanism will fail to work.
Problem 1: tests often fail in pushbuild with a timeout when waiting
for the slave to start/stop/receive error.
Fix 1: Updated the wait_for_slave_* macros in the following way:
- The timeout is increased by a factor ten
- Refactored the macros so that wait_for_slave_param does the work for
the other macros.
Problem 2: Tests are often incorrectly written, lacking a
source include/wait_for_slave_to_[start|stop].inc.
Fix 2: Improved the chance to get it right by adding
include/start_slave.inc and include/stop_slave.inc, and updated tests
to use these.
Problem 3: The the built-in test language command
wait_for_slave_to_stop is a misnomer (does not wait for the slave io
thread) and does not give as much debug info in case of failure as
the otherwise equivalent macro
source include/wait_for_slave_sql_to_stop.inc
Fix 3: Replaced all calls to the built-in command by a call to the
macro.
Problem 4: Some, but not all, of the wait_for_slave_* macros had an
implicit connection slave. This made some tests confusing to read,
and made it more difficult to use the macro in circular replication
scenarios, where the connection named master needs to wait.
Fix 4: Removed the implicit connection slave from all
wait_for_slave_* macros, and updated tests to use an explicit
connection slave where necessary.
Problem 5: The macros wait_slave_status.inc and wait_show_pattern.inc
were unused. Moreover, using them is difficult and error-prone.
Fix 5: remove these macros.
Problem 6: log_bin_trust_function_creators_basic failed when running
tests because it assumed @@global.log_bin_trust_function_creators=1,
and some tests modified this variable without resetting it to its
original value.
Fix 6: All tests that use this variable have been updated so that
they reset the value at end of test.
