Replication snapshot cleanup

[danas]

We have been using the replication feature for a few weeks now and the extra chatter has been causing the SQL Express db for vcenter to crash every so often. This seems to cause an increase in "hidden" snapshots left over from veeam not finishing a snapshot commit from a replication job. Is there any mechanism built in to handle this issue, or will there be in v8? For the most part the snapshots that show up in the snapshot manager in vcenter eventually get resolved, its the ones that aren't listed there that cause issues. They get big, and the more there are the worse the virtual disk performance gets, the more my users feel it. Hoping for some kind of solution so i don't have to constantly monitor the replication everyday. Thanks!!

Dana

[veremin]

Hi, Dana,

Actually, in the process of creating/deleting snapshots VB&R plays only "requester" role, sending corresponding instructions to VMware to create/consolidate snapshot. Thus, the problem with orphaned snapshots is typically related to VMware rather than to Veeam side.

In general, the issue shows up when the VM original disk is still locked by another process at the time the remove snapshot command is received by the host. The unique identifier of snapshot is removed from the VMware database before the actual snapshot is deleted. As the result, the snapshot exists, but is no longer shown in Snapshot Manager.

Another reason might be high disk I/O latency at the time of consolidation process; it might be worth reviewing this article for better understanding.

In order to remove orphaned snapshots, you can create another snapshot, then perform a “delete all” operation in Snapshot Manager. All snapshots, including the ones not listed in snapshot manager will be consolidated. Also, you can use Veeam ONE to be informed about orphaned snapshots' presence.

Hope this helps.
Thanks.

[Vitaliy S.]

Just want to add that in Veeam ONE you would need to use Custom Infrastructure report to identify orphaned snapshot files. This report uses datastore tasks which crawl through the datastore and identify all files stored on monitored datastores.

To build the report you need, just select Virtual Machine as the desired object type, and then add the required columns (such as Snapshot name, Size, Directory, VM name etc.) to the report. Thanks!

[danas]

I installed VeeamONE yesterday and it does just fine telling me if i need a consolidation or not and if i have too many snapshots hanging around. VMware put in a consolidation feature just to handle this kind of situation with version 5 but its a manual process that can impact the running vm disk IO. So I guess the short answer is buy VeeamONE and setup snapshot alerting? It's a little frustrating because we never had an issue with this before using Veeam for replication. One other question, is it normal for SQL Express to be a bit overwhelmed by the extra chatter from replication? Once a day or so we get a transaction log full error that takes out vCenter for bit (even though the DB is set to simple recovery mode). Would it be worth upgrading to SQL standard? Our shop is pretty small, just 15-20 VMs and 3 hosts. Thanks!

[Vitaliy S.]

So I guess the short answer is buy VeeamONE and setup snapshot alerting? It's a little frustrating because we never had an issue with this before using Veeam for replication.

I have suggested this solution so you could be notified about these issues, but I would still recommend to investigate the initial reason for this snapshot commit issue. You can use Veeam ONE free edition, which also should be sufficient to get alerts for orphaned snapshots.

One other question, is it normal for SQL Express to be a bit overwhelmed by the extra chatter from replication? Once a day or so we get a transaction log full error that takes out vCenter for bit (even though the DB is set to simple recovery mode). Would it be worth upgrading to SQL standard? Our shop is pretty small, just 15-20 VMs and 3 hosts.

Could this error be one of the reasons why vCenter Server can potentially have unfinished tasks for VM snapshot commit operations?