SnapShot Deletion timing out and Killing Servers

[fastpud]

Hi,

I had a problem with my production system last week. I went to delete some large snapshots, (3 of), and the VM failed to restart with the following error

"The parent virtual disk has been modified since the child has been created" . using backup etc I managed to get the server back up after a while.

This morning I tried to replicate the problem in my test environment on a working VM.

1. Created a number of Snapshots

2. Added new drive

3. Extended drive size

4. Created snapshot

5. Copied over 60Gb of data to new drive, making the snapshot file over 60Gb.

6. Deleted all snapshots to merge them into the main VMDK

The test server died with the same error we got on the production server last week. It looks to me that the snapshot deletion process is stopping half way through the process thus leaving the server in an unknown state.

Has anyone came across this issue before? Is it a VM issue or a SAN issue?

Cheers,

Shay

[AntonVZhbankov]

>2. Added new drive

>3. Extended drive size

This is actually very bad practice AFTER taking snapshot.

---

MCSA, MCTS, VCP, VMware vExpert '2009

http://blog.vadmin.ru

[fastpud]

OK.

I just added that step in to see it it had any bearing. On the production machine, I didn't do that but still lost the VM when deleting the snapshots.

Any other ideas?

Shay

[AntonVZhbankov · Accepted solution]

I had VMs with BIG snapshots commited, I have VMs with big number of snapshots commited. All went ok.

Don't forget that you sometimes need a lot of space when you commiting BIG snapshots, and especially several BIG snapshots. Once you started snapshot commitment do not touch VM till commitment finished.

---

MCSA, MCTS, VCP, VMware vExpert '2009

http://blog.vadmin.ru

[fastpud]

Is there a rule of thumb for space requirments eg 1.5 times the total snapshot size?

Shay

[AntonVZhbankov]

A rule of the thumb is "don't use snapshots that lives for long time on production VMs unless it's absolutely required". For exmaple last version Veeam introduced in their Backup & Replication such feature as "safe snapshots". Since all VM backups start from taking temporary snapshot, it check for delta file growth and don't allow delta to grow more than 100 MB, so any snapshot commitment would be pretty fast and won't take a lot of disk space.

If you want to commit multiple snapshots, and there are BIG snapshots make sure that there is enough space for all snapshots to be 100% size. Or commit snapshots one-by-one - it would take more time, but require less space.

---

MCSA, MCTS, VCP, VMware vExpert '2009

http://blog.vadmin.ru