February 21, 2006 at 6:33 am
Something strange recently happened to one of our snapshot replication publications. The snapshot agent for the publication was consistently completing in about 11 and a half minutes. All of a sudden it took almost 19 hours to complete and that was the last time it was successful. That was about a week ago and since then we have been stopping the agent once it starts to interfer with daily processing. I don't understand how it all of a sudden went from under 12 minutes to multiple hours. Last night I monitored it for about an hour and a half before it started causing locking issues and I stopped it. The only thing that happened from the last day it completed normally to the next when it took close to 19 hours was that day the publisher, distributor and subscriber instances all were restarted a couple of times. I'm not sure what caused the restarts, I believe there were server problems. The servers are in Europe, but I help support the databases as they used to be local, but were migrated to Europe last year. Does anyone have any ideas what could cause this? The DBAs and application support groups in Europe insist no changes have been made, but something happened, I just don't know how to figure out what. I am thinking about dropping and recreating the snapshot publication. Would that be a good idea?
February 22, 2006 at 2:03 pm
I dropped and recreated the snapshot publication. Last night the replication ran normally and the snapshot agent took about 13 minutes to complete. The servers also went through a clean reboot last night, so I'm not sure if it was fixed by recreating the publication or the reboot. I also found out that when the servers were restarted when the replication problem started it was because of power testing at the data center and instead of reboots the servers were basically just unplugged. They are now reporting problems with the transactional replication also on the servers, it seems that the databases are not in synch even though I have continuous replication set and find no errors in the logs.
February 22, 2006 at 6:48 pm
due to the "unplugging", replication transactions could have been rolled back
i had experienced it before, 5 hours of getting to commit transactions, when suddenly aborted due to n-number of reasons, rollback occurs more than 3x longer
this might or might not be the case for you...
HTH
February 23, 2006 at 6:49 am
Are you saying the transactions are rolled back at the publisher? I'm not sure that is the issue, since there was a clean reboot the night before last. The problem now is it looks as if the subscriber is not in synch with the publisher in our transactional publication. Is it possible that during the power testing transactions were applied at the publisher, and never made it to the subscriber. I'm afraid we may have to reinitialize the publication, which could take awhile.
Viewing 4 posts - 1 through 3 (of 3 total)
You must be logged in to reply to this topic. Login to reply