Slow performance: checkpoint waits, ZFS pools

• Article Type: General
• Product: Aleph
• Product Version: 20

Description:
We are experiencing extremely slow performance on our production Aleph server. The problem has been with us for over three weeks. It became significantly worse over the weekend - March 3-4 - and continued into today. Our DBAs are attempting to resolve the problem, and have opened a Level 2 service request with Oracle. The database (and Aleph application) was restarted this morning at 09:03 local time.

On the database side we are seeing an extremely high number and duration of "enq: KO - fast start checkpoint" waits.

Resolution:
[From site:] On the Aleph (Solaris) server, one of the administrators noted that ZFS pools can perform poorly for things like database logging activity if they become more than 80% full. There is a technical explanation of this here: http://www.solarisinternals.com/wiki/index.php/ZFS_Best_Practices_Guide#Storage_Pool_Performance_Considerations
We have been running several databases servers on ZFS up to now and had never seen this problem before. I requested additional storage (on the Aleph server) and the checkpoint waits subsided.

• Article last edited: 10/8/2013