Comments on: Update: MySQL tmpdir on tmpfs

By: arun

arun — Mon, 31 Aug 2009 19:58:47 +0000

excellent …. I agree with Mark R … the tmpfs size limit might be an issue

http://dev.mysql.com/doc/refman/4.1/en/server-options.html#option_mysqld_tmpdir

By: Mark R

Mark R — Wed, 24 Jun 2009 07:03:54 +0000

Using tmpdir on a tmpfs is fine provided it doesn’t get exhausted, say you have a 2G tmpfs and occasionally a query needs to do a filesort using 3G of temporary space, this query will fail.

Calculating how much space a filesort() needs is nontrivial and it often seems like a lot more than you expect

By: arjen

arjen — Tue, 16 Jun 2009 22:43:44 +0000

Thanks Harry, I’ll update the post to make sure people don’t trip over that. Wasn’t needed for the setup I was dealing with.

By: arjen

arjen — Tue, 16 Jun 2009 22:42:55 +0000

Yea Doug, that’s a valid use example. Of course it’s preferably to not have DB servers on NFS.

By: Doug Lane

Doug Lane — Tue, 16 Jun 2009 14:12:16 +0000

Suppose you put MySQL in a VM environment where 8GB RAM is available and disk access is via NFS–does tmpdir on tmpfs sound viable and preferable?

By: Harrison Fisk

Harrison Fisk — Tue, 16 Jun 2009 13:35:34 +0000

If you do use a shared memory tmpdir, if you are using replication, you will want to change slave_load_tmpdir on your slave to be a real disk which survives a restart.

The issue is that with statement based binary logging, there are many events which create a file for a replicated LOAD DATA INFILE. If you stop your server after some of these events have occurred, but not all, it will break after you restart.