We make extensive use of replication on our system (which is Sql Server 2012 SP2 Enterprise). We have a publisher, 3 subscribers and a distributor. It is all transactional immediate push.
Lately our replication is running behind many mornings and here are the symptoms:
1. I look on the distributor and the cpus are all pegged at 100%
2. The log reader history shows that for hours it has been doing the following:
"Approximately 8000000 log records have been scanned in pass #4, 0 of which where marked for replication"
Basically, for hours the log reader has been scanning log records. We have our publisher currently set to SIMPLE and the log file is usually "medium" in size - maybe 20-30G.
To solve the problem, I usually do a restart of the services on the distributor and then that sends the cpu down to more normal levels. It will do usually a little more scanning but not hours of it and then replication will almost instantly catch up.
Any idea what would cause this? This is causing production issues almost every night and so any tips on how to debug/solve this would be much appreciated.