Details
Major
Description
Our database server has been running fine for a long time, but this night it crashed.
Leading to the crash we saw a JBD2: bad log_start_commit of the filesystem in syslog, followed by:
Jul 25 07:57:12 vmdb01 mysqld: InnoDB: Warning: a long semaphore wait: Jul 25 07:57:12 vmdb01 mysqld: --Thread 140114759440128 has waited at row0ins.c line 2039 for 241.00 seconds the semaphore: Jul 25 07:57:12 vmdb01 mysqld: X-lock (wait_ex) on RW-latch at 0x7f701da49460 '&block->lock' Jul 25 07:57:12 vmdb01 mysqld: a writer (thread id 140114759440128) has reserved it in mode wait exclusive Jul 25 07:57:12 vmdb01 mysqld: number of readers 1, waiters flag 1, lock_word: ffffffffffffffff Jul 25 07:57:12 vmdb01 mysqld: Last time read locked in file buf0flu.c line 1340 Jul 25 07:57:12 vmdb01 mysqld: Last time write locked in file /home/buildbot/buildbot/build/mariadb-5.5.33/storage/xtradb/row/row0ins.c line 2039 ... Jul 25 08:09:05 vmdb01 mysqld: InnoDB: ###### Diagnostic info printed to the standard error stream Jul 25 08:09:05 vmdb01 mysqld: InnoDB: Error: semaphore wait has lasted > 600 seconds Jul 25 08:09:05 vmdb01 mysqld: InnoDB: We intentionally crash the server, because it appears to be hung. Jul 25 08:09:05 vmdb01 mysqld: 140725 8:09:05 InnoDB: Assertion failure in thread 140114818750208 in file srv0srv.c line 2945
All info obtained from syslog is attached to this issue report. It might be a filesystem issue and we keep investigating, but we would like to exclude that the error originated from a bug or config fault in MariaDB.
Gliffy Diagrams
Attachments
Activity
- All
- Comments
- Work Log
- History
- Activity
- Transitions
Can you try again with 5.5.39 ? It should contain improvement to diagnostics and help us to resolve the issue. Your syslog contains also other assertions, is your HD in working order ?
Thank you for your reply, we will schedule a MariaDB upgrade and see... Until now this was a one time issue and of course it would be best if it remained that way 
As I understand, Jan Lindström is currently working on improving diagnostics in case of long semaphore waits, maybe it will help to find the reason of the problem if it re-occurs.