2013/6/12 walter harms <wha...@bfs.de> > > Hi list, > i am trying to understand the incredible use of filepointers in our mysql > server (5.1.53). > under normal condition the server reports 10k-15k open files pointer. > I run a 'flush tables' every 2h to avoid problems, the number of > users/connections is constant > It is an automatic system but a few human users, a happy mix of myisam and > innodb tables running > with no problems on mysqld 5.0 . But now sometimes i am > hit with an insane increase hitting the ceiling at 60k. I do not like the > idea to increase the > limit further because i do > not understand why this is happening (I seems to happen at random times). > > I am running out of idea what my cause the unexpected increase, any idea > what to watch ?
I had a similar problem some years ago with 5.0.84. It was a database with thousand of tables (mostly in MyISAM). It turned to be the --myisam-recover option in Debian init scripts ( /usr/share/mysql/debian-start.inc.sh) . Obviously it's not a good idea to remove it, as if the server crashes, you will needed it. This used to be our db server after getting started: dXX:~# find /proc/`cat /var/run/mysqld/mysqld.pid`/fd -follow -type f 2>/dev/null | wc -l 116810 This is what I saw at the time after removing that MyISAM check: dXX:~# find /proc/`cat /var/run/mysqld/mysqld.pid`/fd -follow -type f 2>/dev/null | wc -l 10730 I would not recommend to keep this as a for-good-solution but a temporary one till you decide what to do. We migrated most of the tables to InnoDB (there was no reason to keep them as MyISAM, it was just legacy stuff). And these problems were gone. I don't know if this can be your case, but this is what happened to us. Hope this helps. Manuel.
