2015-12-01 01:40 PM
I'd like to know what's going on behind the scenes when I ask VSC to backup a VM. I'm assuming it sets up a VMware snapshot (quiesced?) and then does a snapshot of the volume containing the datastore, but I don't see this discussed anywhere.
Why do I want to know? I want to know what the impacts of a backup are on a running VM. Does the VM "freeze up" for a period of time while it's getting settled down for the backup? My VMs are mostly Windows machines. Does it use a Shadow Copy to stabilize the file system before the snapshot?
2015-12-02 05:56 AM
So at a high level, it basically uses the VC engine to alert VM's that a back up to be taken, once the VM's are in a state for backup, it will trigger a back end snapshot at the NetApp volume, this ensures consistency of VM's.
the state of the machines during the backup depends on the config and capability of the VM - for example if a vm can't be backup up live, it will be paused, but most things can be and this, in my experience (and i'm know vmware guru) is pretty rare use cases.
in terms of the use of VSS within the Windows VM - when vmware alerts the vm, this should trigger the VM to alert the OS that it is about to be backed up, this will trigger VSS so ensure consistency within the VM.
I think this relies on the vmware services to be installed on the VM - so it can be triggered.
there are some configuration considerations etc, but at a high level basically VSC triggers a range of vmware activities to make the machines consistent ahead of triggering a NetApp snapshot, to ensure consistent and reliable snapshots.
hope that help.
look up the VSC guide on the documentation site, that may well go into more detail.
a lot of the process may change dependent on vmware versions and VVOLS changes it again from what i understand
2015-12-02 12:59 PM
Although not directly asked/answered, I believe this response may answer your question:
<stolen from NGS>
"To make the VMware snapshot process work the ESX host sends a signal to the guest OS via VMware tools (so VMware tools must be installed, up to date, and working) and then VMware tools triggers the VSS or SYNC service within the guest OS (so that also must be working). Completing the VSS process requires that sufficient memory resources are allocated in the VM (and available in the ESX host) to hold pending writes in memory for the creation of the VMware snapshot. Disk IO performance is also a factor since the ‘quiesce’ process involves sending all pending writes to disk. Both disk IO and memory factors are possible causes/solutions in cases where the quiesce operation simply can’t complete within the specified timeout."
</stolen from NGS>