memory overcommiting

[bolddefender]

I have a 2 host esx 4 cluster with 16GB RAM in each host, memory usage is 11Gb and 13GB. Most of the guests memory are under 10%. eg my largest guest is using 7760MB of host memory but reporting only using 3% guest memory.

I need to put each host in maintenace mode and push some updates to each and reboot the host with out taking down the guests. I thought if i put one into maintenace mode it should over commit on the memory and let all the guests run on one host. Its seems not. I tried this and it shut down a guest or two so I quickly came out of maint mode and boot the guest again.

Is there something I havn't configured right or am I using to much memory.

Thanks in advance.

[NuggetGTR]

Putting a host in maintenance mode should not shutdown virtual machines, that seams strange.

I would think it should be able to move them over with over committing unless you have a reservation on the virtual machines that it cant be met, but still should not shut it down just throw an error.

[mittim12]

I have never seen a guest shut down due to a lack of resources on a host.  At the very  most you would expect the host to simply start paging out the VM's memory if it ran out of physical memory.  That would result in terrible performance but I wouldn't think it would shutdown the guest.

Did you check the event logs of the guest to see if you could determine an exact reason for the shutdown?

[ChrisDearden]

If you had Guests shutting down it sounds more like HA kicked in ? did the hosts enter maintenance mode sucessfully. Did you have strict admission control running ?

[bolddefender]

Sorry for not updating this thread as I'd been busy with other stuff.

I checked the logs for the 2 machines that shut down but the VMware.log for the time period started at the time they were powered on again. The previous log stops weeks before. So it looks like when the machines were powered on they overwrote the VMware.log rather than rename it to the next number in sequence.

I also checked the vmkernel logs on the host put into maintenance mode and the host left running and didn't see anything that jumped about about those two machines.

Anyway its been running OK since and I'll be more careful in the future.