Hi,
We are running ESX 3.5 on 2 hosts, and each host has 1 VM - a total of 2 VMs. The hosts have 24GB RAM each. We have allocated the full RAM allocation, minus the required VMWare overhead, so around 23.5GB, to each VM, and have set up reservations of around 12GB to allow for HA to squeeze both VMs onto one host in the event of a host failure.
We're using VCB backup to handle snapshot-level backups of our 2 VMs on a daily basis, and the backup is failing, and not deleting the snapshot, for one of our VMs, which is causing it to grow and eventually fill the SAN storage allocation. Our hosting provider, who manages the ESX cluster, is blaming our RAM allocation for this - they're maintaining that we should maintain at least 30% of the total host RAM on each machine free to allow VCB to function. The VCB proxy itself is running on a separate machine outside the cluster, and outside of our control.
Can anyone confirm this? Does VCB really require substantial quantities of free RAM, on the actual ESX host machine running the VMs it's backing up, to function? Does VMWare need lots of free RAM to create a snapshot? We are getting the snapshot created, though how long it's taking I'm not sure, but I think the problem is with VCB not the actual creation of the snapshot, but I may be wrong there.
Also, we've left around 16% free disk space on the SAN cluster, which is more than enough space for the delta on the short lifetime of the daily snapshots. The system has run for nearly a week with a growing snapshot in a recent failure scenario, until we were informed of the problem. Is there any need for additional disk space beyond the anticipated growth size of the snapshot, on the SAN itself? Again, the VCB is backing up to an external disk storage array which isn't part of the SAN storage allocation we have.
Any thoughts would be much appreciated!
Regards,
James
