Bug #45956 | Performance schema: exotic mysqld startup, bad results | ||
---|---|---|---|
Submitted: | 6 Jul 2009 0:03 | Modified: | 14 Jan 2010 18:07 |
Reporter: | Trudy Pelzer | Email Updates: | |
Status: | Closed | Impact on me: | |
Category: | MySQL Server: Performance Schema | Severity: | S3 (Non-critical) |
Version: | 6.0.12-alpha | OS: | Any |
Assigned to: | Marc ALFF | CPU Architecture: | Any |
[6 Jul 2009 0:03]
Trudy Pelzer
[7 Jul 2009 14:16]
Marc ALFF
See related Bug#43533
[15 Jul 2009 18:27]
MySQL Verification Team
Thank you for the bug report. Verified as described: miguel@lara:~/dbs/6.0pfs$ libexec/mysqld --port=3307 --performance_schema_events_waits_history_long_size=1000000000000000 090715 15:24:36 [Warning] Buffered warnings: option 'performance_schema_events_waits_history_long_size': unsigned value 1000000000000000 adjusted to 4294967295 090715 15:24:36 InnoDB: Started; log sequence number 0 46409 090715 15:24:37 [Note] Event Scheduler: Loaded 0 events 090715 15:24:37 [Note] libexec/mysqld: ready for connections. Version: '6.0.12-alpha-debug-perfschema' socket: '/tmp/mysql.sock' port: 3307 Source distribution 090715 15:25:17 [Note] Got signal 15 to shutdown mysqld 090715 15:25:17 [Note] libexec/mysqld: Normal shutdown 090715 15:25:17 [Note] Event Scheduler: Purging the queue. 0 events 090715 15:25:18 InnoDB: Starting shutdown... 090715 15:25:19 InnoDB: Shutdown completed; log sequence number 0 46409 090715 15:25:19 [Warning] Forcing shutdown of 1 plugins 090715 15:25:19 [Note] libexec/mysqld: Shutdown complete safe_mutex: Trying to lock unitialized mutex at safemalloc.c, line 520 090715 15:25:19 - mysqld got signal 6 ;
[20 Oct 2009 0:10]
Marc ALFF
Please retest using mysql-trunk-perfschema, which is the performance schema port for 5.5. Changing the report to 'Need Feedback'.
[20 Oct 2009 17:34]
Peter Gulutzan
With mysql-trunk-perfschema pulled today ... (1) fixed (2) fixed, although on shutdown I see: Error in my_thread_global_end(): 3 threads didn't exit (3) fixed (4) not fixed, now mysqld won't start
[4 Nov 2009 0:02]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/commits/89235 2936 Marc Alff 2009-11-03 Bug#45956 Performance schema: exotic mysqld startup, bad results When initialization of the performance schema failed, the server was internally in disabled state, but show variables like "performance_schema" continued to report "ON", which is misleading. With this fix, performance_schema=OFF when initialization fails internally. Also, more warning/information messages are printed in the server log, to explain if/why the performance schema is enabled or disabled at startup.
[10 Nov 2009 18:07]
Marc ALFF
Documenting previous findings (4) not fixed, now mysqld won't start In fact, the server *did* start, but starting the server took several minutes. This is because mysqld needed to allocate 3.2 Gb of memory on a 4 Gb RAM laptop, which caused extreme paging and swapping. The machine was completely killed by mysqld, and unresponsive. The reason for allocating this memory was: - max-file-classes=100000 (set in the test) - max-thread-instances=1000 (by default) To maintain per thread per instrument class aggregates, the server tried to allocate 1000 * 100000 records, or 100 Millions records for file classes aggregates alone, used for the table EVENTS_WAITS_SUMMARY_BY_THREAD_BY_EVENT_NAME. All the sizing parameters now have practical limits to avoid this misuse.
[10 Nov 2009 20:59]
Marc ALFF
Queued in mysql-trunk-perfschema (5.5)
[14 Jan 2010 10:05]
Marc ALFF
Merged in: - mysql-next-mr (Celosia / 5.5.99-m3) - mysql-6.0-codebase (6.0.14)
[14 Jan 2010 18:07]
Paul DuBois
Not in any released version. No changelog entry needed.