SInce you are already dumping all VM's to a list (/tmp/vms_list) then comparing that list with the input, list why not use that loop to check for the existance of a file in the VM directory. If it exists add it to the /tmp/vms_list if it doesn't then don't. Then use the /tmp/vms_list to perform the backup. As part of the generate the /tmp/vms_list loop do all the checking for raw ot snapshots rather than burry them in the backup loop. In the generate /tmp/vms_list loop you can check for the existance of different files. The existance of a daily file would add that VM to the /tmp/vms_list every day. The existance of a weekly file would add that VM to the /tmp/vms_list only on friday or whatever.

Thanks for listening.

Kicking off different jobs from a Linux server probably makes more sense or using multiple cron jobs. Locating the input files on an NFS mount would make them easily accessable without needing to directly access the ESXi host.

Thanks again.

However, I have yet another lame newbie question. Well, sorry, but here it comes.

Setup: 2 HP servers running embedded ESXi and hosting the VMs locally. Both servers have enough capacity on the datastore to host the VMs from the other server.

Question: Would it be possible to add the datastore of the other server in both servers and kind of cross-backup them? VMs from server 1 -> server 2 and VMs from server 2 -> server 1. This would have been done daily scheduled by crontab if possible.

Pros: If a server go down I can power on the VM straight away on the other physical server without having to copy it across the network first (slow due to big VMs). Easy testing of backups since they´re allready on a datastore where they can be booted and tested directly (without network connection for the tested VM).

This would be sufficient backup of the VMs since the most important data form within the VMs are backuped to a tape device anyhow. I am pretty much not allowed ANY investments right now due to the economic state of the company, so network storage or a backup server is kind of hard to get funding for right now. My second option is using an old server with server 2003 OS and USB attached drives. The scheduling through windows seems like the easiest solution for me anyhow. Might use both solutions with a backup weekly to the old server just in case. Transfer to/from the old server would probably be too slow to use it for daily backups.

Unless I get a negative response I will test this setup. I suppose it depends if it´s possible to add the datastore of the opposite server or not. I´ll at least give it a shot.

Thanks for taking the time to write such a money-saving script for the rest of us

I'm guessing sync does a few extra things like resend packets, manage failures etc - just a guess. I've been wanted to read up on it for a while just never enough time All I know is that with async set my backups are about 5 times quicker. If you need some timings to compare post some of yours and i'll post some back.

Cheers