Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

I have about 20+ Ubuntu Guests running on Vmware Server 2.0.2 and every night a backup script takes a snapshot of all running VMs, backs them up, then removes the snapshots.

Every morning I check the machines to find at least 1 or 2 are somehow corrupted. It can happen in guests running Ubuntu Desktop and Ubuntu Servers in versions 8.04 through 9.04. Here is what I mean by corrupted:

1) Postfix dies

2) Zarafa server dies

3) cron jobs on web servers die

4) Firefox dies

5) Gnome Desktop dies

If I reboot them then everything goes back to normal. However, it's causing a LOT of disruption. Are snapshots just flaky? I hate to shudown all the VMs in order to do a nighly backup, but I will if I have to.

share|improve this question
3  
I would say that you're pushing VMWare Server a bit too much to be honest, you're well into ESX terratory with 20 guests and the snapping you're doing. –  Chopper3 Jan 21 '10 at 14:42
    
What means "dies"? Do you have any related logs in /var/log/messages. Do you see oom-killer? If so may be you are overcommiting the memory too much. –  Mircea Vutcovici Jan 21 '10 at 14:55
add comment

2 Answers

Are you sure your understanding of VMWare Server's snapshot tool is what you think it is?

The snapshot tool that ships with VMWare Server is intended to allow the rolling back of changes, not backups.

i.e. Once you create a VMWare Server snapshot you can't just delete it (unless you want to lose changes), you have to merge it back into the VMDK.

It sounds like what you are doing is creating a snapshot and backing up the primary VMDK. During this time the guest is still running and writing changes to the snapshot file. You are then deleting this snapshot, which is throwing everything out because your disk is being reset to when you started the backup.

Given that you are unintentionally shooting your servers in the head each night, here's a few things you should do as soon as possible:

  1. Only run a couple of guest VMs on VMWare Server - it isn't meant to run 5+.
  2. Don't use snapshots in VMWare Server, they are slow and you can't get rid of them easily.
  3. Evaluate and deploy ESX. It has tools for creating backups the way you want (at a price).

If you have to run VMWare Server then forget about live, VM-level backups. Look at doing data-level backups and then every few months shutdown your guests and make external backups.

share|improve this answer
add comment

Personally, with VMware Server (or Workstation or Fusion), I always suspend or shutdown my VMs before doing any snapshots since all write activity (such as virtual memory swapping, or application logging) is a change to the VM.

To second @Chopper3's comment... you really should be moving to a "real" virtualization tool now: ESX or ESXi for that many guests.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.