ASM/ME Smart Copy - prevent Linux VM from Saving memory to disk?

Question

Hi, I have a cluster of two Windows Server 2012 Standard Edition servers running Hyper-V, with a Cluster Shared Volume (CSV) that lives on an EqualLogic PS6500E SAN.  I've configured the EqualLogic HIT/ME 4.5 tools on both servers, and can perform scheduled Smart Copies of a collection consisting of all of the VMs on the CSV.  There are eight VMs now, five of which are Scientific Linux 6.3 or 6.4, and the other three are various versions of Windows. When I perform a Smart Copy, the Windows VMs are aware of the VSS request, they tell their applications to pause for a snapshot, and happily continue running throughout the Smart Copy.  However, the Linux VMs (which are not aware of VSS) can't do this, so they save their RAM to disk and enter a paused 'Saved' state.  Each VM server puts its Linux VMs to sleep one at a time, so depending on whether the VMs are spread over the two servers, or concentrated on one single server, it ends up taking up to four minutes to perform the entire Smart Copy, including resuming from the Saved state. The Linux VMs are unreachable on the network for pretty much the entire duration.  Any SSH connections to these VMs are likely to time out if TCP Keep Alive packets are used, which is usually the default in most SSH clients.  Even if I disable Keep Alive, the session still times out and closes if I happen to try typing during the Smart Copy window. We may have people from around the world logged into the server at any given time, so it's difficult to find a best time to perform snapshots that will minimize disconnects.  I'd prefer not to have to explain to people that they need to turn off Keep Alive just for these servers, or to suggest tools like Screen to keep a detachable session running (though that is a wonderful tool). Is it possible to disable the saving of VM RAM to disk and the pausing of these VMs?  I believe the ASM/VE version has this ability, but I can't find it in ASM/ME. Thank you!

bobthebrewer · Answer

I haven't opened a case yet, thought it might be quicker to start with the community.

Doing this would defeat the purpose of using ASM/ME, though, wouldn't it? I know there's no VSS support in Linux, but it's nice to be able to take one single volume snapshot and know that all of the VMs in it are crash-consistent. All I really want is to eliminate the long save/restore of the Linux VM's RAM.

FluidFS

Was this post helpful?