We have a relatively busy site with ca. 30.000 connections per day. Things run fine, except that 1 or 2 times per day the MySql server crashes for no apparent reason and then restarts automatically. In the error logfile we see that it says that some tables are marked as crashed, but actually they have not, they are all still fine and usable. In the logfile I have added here, it can be seen that we have twice a strange 'Got timeout reading communication packets' from suspicious sources, but they are 5 hours, respectively about 1 hour before the actual crash happens at 06:25.10. The last message with a crashed table is seen 28 minutes before the crash happens, so its probably not related? Could it be that one of the 2 warnings that can be seen is triggering the crash?
Best Answer
This sort of thing is typically caused by an out-of-memory error. It could be a segfault/bug in MySQL, but this is less likely. Check your syslog for additional clues, there should be something in there, regardless of whether it is an OOM or a segfault.