Deleting snapshots

ashman

Gawd
Joined
Mar 28, 2011
Messages
811
So I am deleting snapshots on one of my clients ESXi hosts, running 4.0, snapshots the system made that shouldn't be there. I start the process and at the time I started it the host had about 81GB of free space, low I know, but good enough. As the process goes it starts eating up free disk space to the point where I am beginning to panic because I don't know how far its going to go. Fortunately it stopped eating up free disk space which tells me its merging the snapshots and will delete them soon. I am currently holding at 44GB of free space.....

Anyone experienced this before?
Can it eat up so much disk space that it could crash the host, or would it not start the process if there wasn't enough free disk space to begin with?

Thanks

AM
 
If you are thin provisioned, it will consume more space in the base disk as the snapshot is merged, then you will regain space once the snapshot commitment is completed and the delta file is deleted.

If the system is very large, or if there are issues (disk consolidation needed, etc.), I always find a viable maintenance window to shut down the VM, check to ensure all files are writeable (i.e., not locked by another process or VM), and then perform the snapshot. I like to SSH onto the host to view the virtual machine's underlying files in real-time, and also ensure that everything is committed and purged properly.

One of the best and worst features of virtualization (and enterprise storage in general) are snapshots. They can help and hurt in such significant ways if not managed effectively.
 
From what I have described, does it sound like its merging the snapshots now since its not consuming any more free disk space?
 
Not to try and hijack the thread but your mentioning of consolidation has me wondering: What's the best practice with regards to consolidating and snapshots? Consolidate, then snap (which is what I have been doing)?

As for the OP's question, I'm not an expert by any stretch of the imagination, but based on what I have seen in my system, yes, Once the snaps have been merged you will not consume any more space. And as REDYOUCH mentioned, you may even get some space back when all is said and done.

Bill W
 
Thanks, it finally stopped consuming disk space and did merge and delete the old snapshots, I eventually wound up with 126GB of free space, so its all good, thanks for everyone's help.

Wildbill, why are you manually taking snapshots? They are not best practice for backups...
 
Not to try and hijack the thread but your mentioning of consolidation has me wondering: What's the best practice with regards to consolidating and snapshots? Consolidate, then snap (which is what I have been doing)?

As for the OP's question, I'm not an expert by any stretch of the imagination, but based on what I have seen in my system, yes, Once the snaps have been merged you will not consume any more space. And as REDYOUCH mentioned, you may even get some space back when all is said and done.

Bill W

Thanks, it finally stopped consuming disk space and did merge and delete the old snapshots, I eventually wound up with 126GB of free space, so its all good, thanks for everyone's help.

Wildbill, why are you manually taking snapshots? They are not best practice for backups...

I don't think WildBill was using snapshots as backups, just wondering about the different between deleting snapshots and "consolidate". Deleting the snapshot will merge the delta file with the underlying base disk (or drop the delta file depending on "you are here"). Consolidate will manually merge all snapshot delta files with the underlying base disk, which is helpful if there is a failed snapshot commitment, or locked delta (or underlying base disk) file. Obviously, being in any situation where you need to consolidate is not good, but sometimes happens with VADP-based backups (like EMC Networker). When this has occurred in the past, I have shut down the machine, checked for locked file ("touch *"), and then consolidated. Please keep in mind that it can be time consuming, especially if you take into consideration the size of the deltas and how long they have lingered.
 
Someone correct me if I'm wrong but my understanding was as it merges the delta file with the VM you roughly need VM size + (delta file size * 2) because it doesn't delete the delta file until after the merge has completed and "potentially" you could be adding the entire delta file to the size of your VM. If I find a snapshot that has been around (run weekly reports to ensure this doesn't happen) I check the delta file size, the VM size and how much space is available before I delete the snapshot.
 
Someone correct me if I'm wrong but my understanding was as it merges the delta file with the VM you roughly need VM size + (delta file size * 2) because it doesn't delete the delta file until after the merge has completed and "potentially" you could be adding the entire delta file to the size of your VM. If I find a snapshot that has been around (run weekly reports to ensure this doesn't happen) I check the delta file size, the VM size and how much space is available before I delete the snapshot.

This is correct for Thin-provisioned disks. For Thick-provisioned, it would be VM size + (delta file size * 1), as the maximum size of base with delta committed could never exceed the base.
 
This is correct for Thin-provisioned disks. For Thick-provisioned, it would be VM size + (delta file size * 1), as the maximum size of base with delta committed could never exceed the base.

except in old versions like taht we merged from the top down, so you theoretically needed:

Base + (minimum:(sum of all snapshots, base * number of snapshots)).
 
Back
Top