Snapshot fails unless we snapshot through vSphere first

[moniker131]

We have some severs that we snapshot once a week. They fail when the job runs. If I re-run the job on our off night, it still fails (this tells me this isn't any sort of load problem). If I take a snapshot via vSphere client, and remove it, and then run the job, it works. One thing that stands out to me is after the manual snapshot, the next snapshot (done by backup exec) is much quicker. This is probably because the data is already quiesced. Is there a snapshot timeout that we're hitting when the job normally runs?

[Colin_Weaver]

Not exactly sure what your comparison is but some thoughts

Once Backup Exec's Agent for Windows is installed inside a VM a modification is made to the pre-freeze and post freeze script files relating to VSS snapshot providers within the OS of the VM.

These script files are used whenever a vSphere Snapshot is requested (or deleted) as a call is made through the VMware tools which calls the scripts - it does not make any diffence if Backup Exec initiates the vSphere snapshot or if another product does it, the change due to the install of the Backup Exec remote agent is still made (and then undone when the vSphere snapshot is removed)

This has been known to cause problems if you mix:

- Backup Exec vSphere Backups

- 3rd Party requested vSphere Snapshot changes

- Traditional agent backups of the same VM using Backup Exec.

the problems might affect CBT (so incremental backups) and / or or cause outright failures - as such some examples

1) if a traditional agent backup of the VM is running, then any call to create or remove a vSphere triggered snapshot at the same time will cause the traditional agent backup to fail.

2) Extended chains of snapshots may cause delays that result in a timeout (irrespective of other load on the server)

3) Incremental backups may run as  fulls (or fail depending on settings) if CBT is unexpectedly changed between the full and incremental backups.

You do therefore have to take care with scheduling etc to avoid any conflicts or any breaks in the sequences of incrementals back to the full.

Note: As far as I am aware, VMware don't recommend maintaining long running snapshots due to the potential for adverse effects on performance and suggest these are only used for backout plans for change control practices or similar.

[moniker131]

Thanks for the reply, Colin.

I did read about those scripts when researching this issue. Since a manual snapshot through vSphere works, I was assuming those scripts are working fine since like you said, they run no matter what initiated the snapshot.

1. The remote agent is not installed on these servers, vm snapshot backup only.

2. We do not have any pre-existing snapshots on the server. Just the temporary snapshot while backing up.

3. We only do weekly fulls on this so no full/inc chain to worry about.

[Ben-Veritas · Accepted solution]

Try setting the registry keys from this technote to see if that resolves the issue.  This will allow BE to retry taking the snapshot a few times before failing.

https://www.veritas.com/support/en_US/article.100028133

Note the reg key in step for might be HKEY_LOCAL_MACHINE\SOFTWARE\Symantec intead of Veritas, depending on when and what version of BE you have installed.

[moniker131]

This has helped. Instead of failing, these jobs are now taking "crash consisten" backups. Thanks.