I have a 2 data/mysqld node cluster with 1 fairly good sized database (4 gb). I recently move one of the larges tables to disk tablespace as it's memory usage on the data nodes was growing exponentially.
The database serves an apache2 web application and once in a while(weekly) the mysqld process locks, which in turn Zombies every apache session on the server. I can do a
kill -p (mysqld PID) and it instantly retries the transactions and all the sessions complete with no issue. Trouble is, I can't find anything in any apache, kernel, system, or other log that indicates any issue. Even the mysql trace files just stop. It seems like it locks before it has a chance to log anything. Any good place to start looking? the data nodes don't miss a step and mysqld restarts almost instantly after its killed.
The database serves an apache2 web application and once in a while(weekly) the mysqld process locks, which in turn Zombies every apache session on the server. I can do a
kill -p (mysqld PID) and it instantly retries the transactions and all the sessions complete with no issue. Trouble is, I can't find anything in any apache, kernel, system, or other log that indicates any issue. Even the mysql trace files just stop. It seems like it locks before it has a chance to log anything. Any good place to start looking? the data nodes don't miss a step and mysqld restarts almost instantly after its killed.