Snapshot question

[vmproteau]

We don't use snapshots often and if we do they don't remain for more than a day.

I try to leave room on my datastores but, we've had engineers push them to the edge. What exactly would happen if a datastore containing an active growing snapshot ran out of disk space. This sounds like a bad scenerio.

[mcowger]

SCSI write errors would be returned back to the VMs on the datastore. The guest OS's would deal with them per their design...most linux's would set the disk read only - not sure what windows would do.

--Matt

VCP, vExpert, Unix Geek

[anoni22]

It is... This happened to us once and it brought down the VM with the snapshot and all the other VM's on that LUN. If I'm not mistaken, we also had to restore a few of those VM's from backup from that incident

[Chamon · Accepted solution]

I believe you would crash the LUN and not have access to it any longer. You should always leave 10% of freespace on the datastore. I haven't done it but heard of a vRanger backup snapshot filling a LUN and the VMs were no longer accessable.

[vmproteau]

Sounds like there are a number of possible results...all not good. I may do some tests in the lab. I'd like to find clues to recognize it quickly when I see symptoms.

One related general question about snapshots...include memory or not? It seems if you don't include memory in the snapshot, the amount of space used up (at least initially) is greatly reduced. It just seems that if your snapping a powered on machine, you don't really have a choice except to include memory. If you snapped a live machine and didn't include memory, wouldn't reverting back have some problems for the OS? running applications?