Execute pre-freeze and post-thaw
-
Using webhooks, if I understand correctly, requires installing a Node.js server. This server would then call the commands to put Oracle into backup mode.
This solution is cumbersome because it would require maintaining Node.js.
Is there no way to directly use Guest Tools to call a script before and after snapshot, like with VMware or QEMU? If not, is this planned?
-
@jsajous26 AFAIK this is not possible to call a script inside the VM through the guest tools. Maybe on veeam side there are other way to do this ( throug ssh / powershell) ?
-
You can read key/values from the xenstore, and write some (from VM to outside), but you cannot write values "in live" from outside the VM to the inside.
-
@olivierlambert do the guest tools quiesce the filesystems before snapshotting?
-
Tools are aware of a snapshot so you don't have blocks in flight.
-
@olivierlambert So there is information coming from outside into the VM
Is there a particular reason not to implement freeze/thaw scripts in that case? The tools know there is a snap taking place and you said that you can write info from VM to outside, which would have to happen to tell the system that blocks have been committed and is ok to start. Seems like the communication is there, just need to implement the scripts.
Coming from VMware land, this is expected behavior of the guest tools.
I agree with OP, the webhooks looks janky, does not appear to be in the main docs, and the linked docs says it is unstable after 3 years.
-
No, it's not using the same mechanism. It's a PV call (if I remember, but I'm not sure). You cannot write a xenstore key in live to a VM (only on boot).
-
And beyond that, I'm not sure we actually want this behaviour from a security perspective. But it's another topic.
-
@olivierlambert ok, but it wouldn't have to. The guest tools always run the scripts (if they exist) when it sees a snap happening and then reports out the status the same way it says that the block commits are complete.
-
They aren't meant for that, I have no idea how difficult it would be to make the PV drivers to actually "do something" (what exactly? write a value in the xenstore?). Maybe @teddyastie could provide an opinion on that.
-
You can read key/values from the xenstore, and write some (from VM to outside), but you cannot write values "in live" from outside the VM to the inside.
It is, but XAPI doesn't provide a interface for it.
do the guest tools quiesce the filesystems before snapshotting?
Tools are aware of a snapshot so you don't have blocks in flight.
do the guest tools quiesce the filesystems before snapshotting?
Guests kernel are aware, as it is them that are performing a "suspend" on toolstack request (thus quiece filesystems); although "tools" can only observe that the system has been suspended after the fact by measuring side effects, and not orchestrate it.
It's because suspend/resume operation doesn't come from "guest tools" actually, but instead from the kernel drivers. So userland tools has no say on it.
-
There used to be quiescent snapshot capabilities in older versions (mainly for Windows VSS support), but it has since been removed. I'd say @Team-XAPI-Network knows more about the reason.
-
@dinhngtu It's possible to run hooks inside the control domain on different stages of VM lifecycle operations, but these need to be installed beforehand, see https://xapi-project.github.io/new-docs/xenopsd/design/hooks/index.html
I think the vm-pre-suspend is the name of the hook the does this, but I can't confirm it. The docuementation is missing the hook, but it exists in the source code: https://github.com/xapi-project/xen-api/blob/master/ocaml/xenopsd/lib/xenops_hooks.ml#L37
-
@psafont Looks like these hooks are for suspend-resume rather than snapshot/backups? I couldn't find any mention of hook calls during the snapshot process
-
@dinhngtu said in Execute pre-freeze and post-thaw:
There used to be quiescent snapshot capabilities in older versions (mainly for Windows VSS support), but it has since been removed. I'd say @Team-XAPI-Network knows more about the reason.
We have vmware with ppdm backups and the vss part is actually causing some annoying issues such as quite long io stalls. But I do understand the reason for vss, especially for applications that aren't crash safe.
-
@dinhngtu Xenopsd does not have a concept of snapshot. Because I understand that "freezing" the VM means suspending it, I think that the suspend hook will get triggered in any snapshot with memory
-
@psafont Freezing/thawing doesn't actually suspend a VM, but it's just a part of the backup process, where I/O operations are quiesced to create a stable view of the on-disk data.
-
Could this feature integrate XCP-ng?
We are looking for a new virtualization solution.
Oracle is a bottleneck for backups.We tested the backup with the Veeam agent, which appears to be successful.
However, replication is not possible for this Oracle point via Veeam. -
What is the purpose of this discussion?
You did create this thread, I think you would know what's the purpose of this discussion
Have you asked VEEAM for a solution, because it seems more VEEAM-related question after all
I have no idea what it means regarding "replication is not possible for this Oracle point via VEEAM". Remember we aren't from VEEAM in here, so you should explain more precisely what do you expect from us to do despite it seems to be a VEEAM limitation? -
@olivierlambert said in Execute pre-freeze and post-thaw:
You did create this thread, I think you would know what's the purpose of this discussion
Have you asked VEEAM for a solution, because it seems more VEEAM-related question after all I have no idea what it means regarding "replication is not possible for this Oracle point via VEEAM". Remember we aren't from VEEAM in here, so you should explain more precisely what do you expect from us to do despite it seems to be a VEEAM limitation?
The problem is not Veeam, but the inability to trigger a pre-free and post-thaw (Quiesce) on a VM before and after snapshot on XCP-NG.
