January 21, 2007 at 11:32 am
I have a situation that looks like a capacity problem with replication but perfmon does not indicate the machine as being anywhere near capacity.
I have several SQL2000 and SQL2005 publishers being pushed through a very large SQL2005 distributor. The distributor was specified to cater for at least 3 times the traffic that it is getting.
Although traffic through the distributor has increased substantially there should still be substantial headroom. The affect is that more transactions are coming in to the distributor than are going out with obvious consequences.
Neither the CPU, memory, disk or network performance counters indicate that the machine is being stressed and yet this problem persists.
If some of the subscriptions are changed to pull subscriptions then the distributor publishes like a Los Angeles porn house and the backlog is cleared very quickly. This emphasises that the sheer number of transactions is not the issue.
If all the subscriptions are switched back to push subscriptions then replication slows to a crawl.
It sounds as if some process involved in replication can only cater for a certain number of transactions on any one machine.
Can anyone shed some light on this?
January 24, 2007 at 8:00 am
This was removed by the editor as SPAM
January 24, 2007 at 12:19 pm
do an sp_who on your distributor and see if there is any blocking
we have a distributor and noticed that the clean up job was running for a very long time and blocking replication. for now we started running alter index a few times a day in the distribution database to see if solves it. another thing we did was schedule the distribution job to only run at night
Viewing 3 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply