scsi unmap trouble

[mrt]

During Veeam jobs that involved applying retention (deleting expired Veeam backup files), I noticed long periods of job inactivity. All jobs would stop processing, yet remain running in a “suspended animation” type of state. After an hour or so, activity would sometimes resume, but other times I would have to resort to rebooting the repository server to get jobs going again (and then deal with the mess it created.) During these periods of inactivity, my backup volume (a FC lun on a newer Compellent) would seem completely dead. I could not even browse its contents with Windows Explorer. No errors or warnings were ever logged by Windows.

After much troubleshooting and lost sleep and elevated blood pressure, I think I finally found the cause: scsi unmap. My repo server is Server 2012 and therefore unmap-enabled by default. When it came time apply retention, Veeam would ask Windows to delete the large old Veeam backup files, at which point unmap would activate and for some reason make the lun unavailable for host i/o for long periods of time.

It’s not hard to find evidence on the internet of unmap choking on large files, but its impact for me was extreme, possibly due the size of the files it was “unmapping.” Has Veeam seen this in the lab? It could simply be that the Compellent handles this feature poorly, but if it’s not, this might be worth a mention in the Known Issues documentation. I’ve disabled unmap in the OS and things have been running normally.

[veremin]

Have you had a chance to open a ticket with our support team?

They will be able take a closer look at corresponding logs and understand the root cause of this issue. If the it’s confirmed that scsi unmap functionality causes this behavior, then, the support team will react accordingly, finding applicable workaround or updating existing knowledge base.

Thanks.