nextlevel
Contributor
Contributor

Delete snapshot speed

Jump to solution

Hi

We have a DELL 6850 4xDualCore 32Gb RAM Server running ESX 3.01.

The disksystem is a PERC 4d RAID with 5x 72Gb 15'000 rpm drives, configured as RAID5

There is a VM with a 12 Gb virtual disk and a snapshot file about 8Gb.

If I commit this snapshot, it took more than an hour complete. The average disk write rate is only about 2'600KBps while this task is running (against 26'000 if I just copy files in a shell)

Is this normal? If yes, why is consolidating so slow?

Any experiences with DELL Raid controllers that perfom better anyway?

Thanks

0 Kudos
1 Solution

Accepted Solutions
virtualdud3
Expert
Expert

Yep, snapshots work great for their intended purpose - to repeatedly return to a given state. Where some people get into trouble is they have a tendency to use snapshots for "long-term backup". Then, the redo logs can grow so large they can fill-up the LUN (although the size of the redo log cannot exceed the size of the virtual disk), or the snapshot won't "apply".

Also, keep in mind that if you delete the "redo log", this does NOT "delete the snapshot". Deleting the redo log will delete any changes made since the snapshot was taken, and the VM will return to its original state (its state at the time the snapshot was taken). You probably knew all of the previous information; however, I have seen quite a few cases where people misunderstood snapshots and so I thought I would pass the information along.

###############

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

############### Under no circumstances are you to award me any points. Thanks!!!

View solution in original post

0 Kudos
7 Replies
virtualdud3
Expert
Expert

This sounds about right.

Applying redo logs (deleting snapshots) can take quite some time to complete. Make sure that you don't stop/cancel the process before completion, as doing so can cause loss of data (as I'm sure you are aware).

So, sounds like you are doing the correct thing and getting a "feel" for the time required.

###############

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

############### Under no circumstances are you to award me any points. Thanks!!!
spex
Expert
Expert

If you delete snapshots your physical disk is written completly random with the information that comes from your snapshotfile. Since the snapshotfile resides in most cases on the same disk there is a lot of competition between read and write. So you can not expect the same speed as in copy jobs that work sequential with large - buffered (EMC: MR5) written blocks.

Regards Spex

nextlevel
Contributor
Contributor

Sounds like you should never let a snapshot grows too much Smiley Sad

I understand the performance difference between sequential and random reads/writes, but 10 times slower ???

However, thank you for keeping me away to buy a faster raid for nothing...

Regards

0 Kudos
Dave_Mishchenko
Immortal
Immortal

The time to commit a snapshot can be a factor of how large the snapshot is plus how busy the VM is as well. Here are some other threads on that. As you mention, it's best to keep the snapshots small and use them for as short a period as possible.

http://communities.vmware.com/thread/74669

0 Kudos
virtualdud3
Expert
Expert

Yep, snapshots work great for their intended purpose - to repeatedly return to a given state. Where some people get into trouble is they have a tendency to use snapshots for "long-term backup". Then, the redo logs can grow so large they can fill-up the LUN (although the size of the redo log cannot exceed the size of the virtual disk), or the snapshot won't "apply".

Also, keep in mind that if you delete the "redo log", this does NOT "delete the snapshot". Deleting the redo log will delete any changes made since the snapshot was taken, and the VM will return to its original state (its state at the time the snapshot was taken). You probably knew all of the previous information; however, I have seen quite a few cases where people misunderstood snapshots and so I thought I would pass the information along.

###############

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

############### Under no circumstances are you to award me any points. Thanks!!!
0 Kudos
nextlevel
Contributor
Contributor

Yep, snapshots work great for their intended purpose - to repeatedly return to a given state. Where some people get into trouble is they have a tendency to use snapshots for "long-term backup". Then, the redo logs can grow so large they can fill-up the LUN (although the size of the redo log cannot exceed the size of the virtual disk), or the snapshot won't "apply".

Uuuuhuuu, good to know ....

Also, keep in mind that if you delete the "redo log", this does NOT "delete the snapshot". Deleting the redo log will delete any changes made since the snapshot was taken, and the VM will return to its original state (its state at the time the snapshot was taken). You probably knew all of the previous information; however, I have seen quite a few cases where people misunderstood snapshots and so I thought I would pass the information along.

Beside, I think the term "DELETE" in the snapshot manager is a miserable choice by VMWARE designer. It is one of the most confusing thing. They should really change it to "COMMIT".

And it would BE VERY NICE TOO, if there were a "Are you sure?", if you revert to a snapshot. More than once accidentaly hit this damn button... Smiley Sad

0 Kudos
dysatguy
Contributor
Contributor

Hello, was wondering if someone could answer a question related to the Delete Snapshots. Since I am new to the VM World and of course have not got any great training thus far on our production ESX 3 server, I created a snapshot before doing an update to our blackberry server that is virtualized. This was a few months ago. I didn't realize that leaving the snapshot not only hampers the performance but can be a problem when deleting them after an extended period of time. I did this and this morning I started the delete of the old snapshots. After the first 15 minutes of Inprogress deleting the snapshot I started to do some reading from people on the forum talking about how this could take hours. To make a long story short my current status is that the snapshot is still committing (I think) but I am not able to log back into the VMWare Infrastructure Client. The message I get is server is taking too long to respond. The other Virtual machines are still running but I cannot get into the client to see the status. I don't even know how big the snapshots on this particular virtual machine are. I guess my question is, is it normal to not be able to get into the client when this process is running (I cannot get into the Http interface either). Next question if the process finishes will I be able to get into the interface again or is there a process I should be restarting. Thanks for your help.

0 Kudos