Unfortunately, we cannot upgrade Postgresql at this time.
We are having delays on our database checkpoint writes of up to 300 seconds. The delays occur only after we delete files from archwal filesystem. We run a script to remove the files all at one time.(up to 3000 files) The command is “find /u04/pgprod/mesprod/9.0/archwal -mtime +3 -exec rm -f {} \;” The other checkpoints occur in milliseconds.
Our Solaris 10 support has recommended changing the zfs filesystem parameter from “primarycache = all” to “primarycache = metadata” since we are using a EMC VNX-5400 SAN as a data store. It is believed that the zfs caching from the delete is delaying the acknowledgement from the VNX or the zfs filesystem when the writes occur.
Are there any adverse issues for postgresql?
Will this resolve the issue with the checkpoint delays?
Will there be any impact to the database specifically how the data is written using fsync?
Thanks and regards,
Bryan Ward
Globitech Inc.