I'm not sure I understand exactly what you're doing. Are you using RDMs and giving each VM a direct LUN to the storage system, or are you presenting datastores via iSCSI? Are you saying you're presenting one datastore per VM?
Managing RDMs for 2500 VMs is simply impractical, and there's a limit to the number of datastores VMWare supports anyway. As for the filesystems, it's true that most filesystems today can survive a hard reboot, but the applications may or may not. It's not 100% guaranteed. Crash consistency provides for the OS to come back, do some filesystem repairs, and hopefully most of your data is intact, but it doesn't guarantee it. For some non-trivial number of user applications, they require a higher level SLA. In our case, it might make sense to just perform standard client-side agent-based backups for those to guarantee restorability, but of course that has its own issues with time and amount of data. In Windows, there are VSS writer processes that control the I/O, and you need to communicate with the VSS writers to tell them to pause while a snapshot is taken to guarantee filesystem integrity (to make sure all I/Os are written and verified). So not only does VMWare need to be aware of the snapshot, but Windows does as well. VMWare Tools allows VMWare to tell the VM, through VSS, to quiesce, and then VMWare can take its snapshot -- it knows to quiesce when it takes its own snapshot. Once that snapshot exists, it's 100% safe for the NetApp (or ZFS) to snap. The added complexity of VMWare and virtualized datastores doesn't help this process. If you constantly hard reboot a busy MS SQL server, chances are it'll eventually not come back the way you want it. Now, with vSphere 6 and vvols, this should be less of an issue because each VM gets a virtual datastore and can be individually controlled, snapshotted, and backed up. -Adam On Wed, Oct 28, 2015 at 6:52 AM, Edward Ned Harvey (lopser) < lop...@nedharvey.com> wrote: > I'm hearing a lot of people here saying "quiesce" the VM, and how many > VM's do you have per volume... I am surprised by both of these. > > What I've always done was to make individual zvol's in ZFS, and export > them over iscsi. Then vmware simply uses that "disk" as the disk for the > VM. Let ZFS do snapshotting, and don't worry about vmware. Every guest OS > (at least every one I've had to deal with) is designed to be able to > survive a power failure (or kernel halt or whatever) so if you ever need to > rollback or restore a ZFS snapshot and reboot the guest, you're effectively > booting that guest as if the power had been interrupted at the time of the > snapshot. > >
_______________________________________________ Tech mailing list Tech@lists.lopsa.org https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
