We have a 2008R2 x64 VM that runs 25 small databases, most are set to full recovery model. We use a Unitrends 822 on the latest OS version release 7.2.0. I recently changed our backup routine from Daily fulls of the SQL Server database to Weekly Fulls with hourly transaction backups.
When I changed the backup plan, I noticed during the transaction log backup, the Unitrends agent would spike the CPU and then start consuming more and more memory. Finally it would spill over into page file and consume it too, ultimately filling up the disk that page file sits on. A reboot of the server is the only way to fix this.
This does not happen on every transaction backup. It is random at best, maybe once or twice a week. When it starts to fail, I see this error message: Failed to queue: Execution of bpr failed, could not request job. Then the CPU spikes and it is just down hill from there.
I have called support and they have tried to fix it 3 times with no resolution. First, it was because "the agent 7.2 was the problem and it could not handle the frequency of the transaction log backup", so we downgraded to 7.0. That didn't fix it. Second, it was because we were running weekly fulls and hourly transaction. We changed it to daily fulls with hourly transaction backups. That didn't fix it. Third, it was because we needed to downgrade to version 6.4 on the agent because the 7 agent was evil. That didn't fix it and made the problem even worse.
I am at my wits end with this and I am about to go Office Space on this thing. Has anyone run into a backup problem on a SQL Server 2008 with many databases using a transaction log backup?