Bug 1840795

Summary: [RFE] Make live snapshot with memory without pausing the guest
Product: [oVirt] vdsm Reporter: Liran Rotenberg <lrotenbe>
Component: CoreAssignee: Liran Rotenberg <lrotenbe>
Status: CLOSED DEFERRED QA Contact: meital avital <mavital>
Severity: low Docs Contact:
Priority: unspecified    
Version: 4.40.13CC: ahadas, bugs, michal.skrivanek
Target Milestone: ---Keywords: FutureFeature, RFE
Target Release: ---Flags: sbonazzo: ovirt-4.5?
pm-rhel: planning_ack?
pm-rhel: devel_ack?
pm-rhel: testing_ack?
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: Enhancement
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2022-04-08 16:48:57 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: Virt RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description Liran Rotenberg 2020-05-27 16:00:39 UTC
Description of problem:
Live snapshot with memory always pausing the VM while making it.
Since in 4.4 we moved to async snapshot, we can prevent the VM pause in the cost of longer convergence and a bigger snapshot disk.

Steps to Reproduce:
1. Create a VM.
2. Run the VM.
3. Create snapshot with memory.

Actual results:
VM is paused during the operation.

Expected results:
VM won't be paused during the operation.

Additional info:
This require us to use VIR_DOMAIN_SNAPSHOT_CREATE_LIVE flag when calling from VDSM to libvirt.

Comment 1 Arik 2020-06-04 16:14:04 UTC
That's something that was discussed back in oVirt 3.3 when memory snapshots were introduced [1].
Besides the above-mentioned downsides of doing it "live", we've realized it would probably make sense to leverage the existing mechanism in VDSM for live-migration - and that was (is?) fairly complex.

[1] https://www.youtube.com/watch?v=xIhPV66uGo8 24:13-25:15 :)

Comment 2 Sandro Bonazzola 2022-03-29 16:16:40 UTC
We are past 4.5.0 feature freeze, please re-target.

Comment 3 Michal Skrivanek 2022-04-08 16:48:57 UTC
no updates for a long time, missed 4.5 GA, closing

Comment 4 Arik 2022-04-10 08:44:01 UTC
This can be an interesting technical challenge but as this option has several downsides to this approach:
1. The snapshot operation could then fail because the 'migration' may not converge
2. The snapshot operation can take significantly more time and during this time the VM is locked exclusively
3. The volume that stores the memory dump can be significantly larger, consuming more space on the storage device
4. It would be hard to predict the size of the volume used for the memory dump so storing it on block devices would be complex
So with no request for this from the field, it has never been prioritized