Hmm, that is peculiar. Mediamanager is managing requests from Media Agents to create/assign new chunk information in the database.
Depending on your configuration, its likely that your synthetic full backups run as single stream, whereas incrementals will use multiple streams and create more load on Media Manager. That could be a significant issue.
I mean, keep progressing with engineering here, and I’m not looking at any logs - but some thoughts:
- If the CommServe is virtualized, get your VM admin to check the CPU ready time. High CPU is often masked by oversubscribed resources on the hypervisor, increasing latency and waiting for a CPU slot to become available to execute a request. Good read
- Pause data aging during your next backup window and see if that helps
- Maybe lower the job watermark or max streams. It may not impact your window all that much if you get higher performance with less streams overall.
Hope that helps!
- It’s physical
- There is a data pruning window during the high cpu times
- Will look into lowering job watermark
We’re also seeing high cpu usage with the SQL service. Devs seem to be focusing their efforts on figuring out what’s causing that to happen. But minimal progress.
Greatly appreciate the response.