Snapshot removal after backup takes a very long time

[KiwiJJ]

Hi,
I have a job that backs up a 280GB VM. After the backup has completed it takes over three hours to remove the snapshot as per below from the logs. Is this normal ?
Should I have "Safe Removal dor Snaphots" selected for the job .... how does this work ? and what is the recommended size ?

thanks,

John

[07.04.2010 23:42:09] <01> Info [VimApi] RemoveSnapshot, type "VirtualMachineSnapshot", ref "/vmfs/volumes/4819489e-315fbc6c-e115-001aa028d1a5/DBMCAAKL02/DBMCAAKL02.vmx-snapshot-74", removeChildren "False"
[08.04.2010 03:08:26] <01> Info [Soap] Loading '112:VirtualMachine' hierarchy
[08.04.2010 03:08:27] <01> Info [Soap] Loaded 6 hierarchy objects from "xxx.xxx.xxx.xxx", 0 Datacenter(s), 1 HostSystem(s), 1 VirtualMachine(s), 3 Datastore(s), 0 ResourcePool(s).
[08.04.2010 03:08:27] <01> Info [Soap] Connection 'xxx.xxx.xxx.xx:443:root:False::0:1' is disposing.
[08.04.2010 03:08:27] <01> Info Task operation: Finalizing target session

[247it]

Hi, I am only new here....but, yes, it can take a while. The snapshot keeps a list of any disk changes since its creation, so if there is a lot of disk activity it will be a lot of changes. Also some software can make more changes than others, even when things appear 'idle', e.g SQL & Exchange log files. Also I have found it can take longer if you already have pre-existing snapshots, so if possible, avoid leaving other snapshots around.

Enabling safe snapshot removal magically prevents a case where the VM becomes unresponsive for a while during the snapshot removal - so it removal is taking 3 hours, I would probably suggest you enable it to prevent the server disappearing from the network during this time.

Tony

[Gostev]

John, the time it takes to remove snapshot depends on its size (how large it has grown for during backup). I would say 3 hours for 280GB VM is waaay too long in all cases, for me this indicate storage issues - like high load or multiple concurrent operations (due to VMFS, SCSI reservations affect operations like snapshot management very significantly).

Does it take this long everyday, or it happened just once? May be you indeed had very large abandoned snapshot on this VM?

Please refer to page 49 of the User Guide for detailed description of the safe snapshot removal feature. As described there, this functionality is designed for pre-ESX3.5 U2 hosts, and in all cases, it cannot improve removal speed of large snapshot - it's purpose is very different. Snapshot will take as much time to remove as it takes ESX to commit all of its data into VMDK files, so it comes down to disk I/O, and no magic can make it happen faster...

Overall, snapshot creation and deletion is handled by VMware for us, we only issue API calls to create/delete snapshots and the rest is handled for us. So snapshot commit operations is not something our product can control, speed up or otherwise affect.

Hope this helps! Thanks!