About a month ago I setup mirroring with our DB, High safety with automatic failover. Ever since the lock waits per minute in the DB went from maybe 2-5 per minute to 22-25 per minute. I am not sure if this was expected or what; but sometimes it spikes even higher than that.
Does this sound off the charts to anyone or normal for mirroring?
Thanks,
Let me know if you need anymore info....
What is the latency between your primary server and the backup server?
This may be normal as transactions aren't commited until after the command is sent to the backup server and the ACK is returned to the primary server. This would cause the locks to be held longer.
|||Latency has be low. If I go back to the last four hours (typical production day) it was 0-5 ms. I have seen it spike more but rarely. Mainly the lock waits were my concern, because they have certainly gone up and stayed up since mirroring started.thanks
|||I should have been more clear. I was asking about network latency from your production server to the backup server and back. IE. Ping from one machine to the other, and multiply by two.|||oh ok, so if I ping from production server to the backup server; all four come back with less than 1ms. And if I go from the backup to the production it is the same.
|||Are your mirroring endpoints are set up with encryption. If you have encryption on and can remove it you will see a ten fold increase in transaction speed in a mirror session.|||There is a database mirroring counter "Transaction Delay" that can help you determine how much latency mirroring is adding to your transactions. Some applications can certainly notice it more than others in terms of lock waits.
No comments:
Post a Comment