CloudStack-KVM integration for Veeam B&R

[yadvr]

Just wanted to share some progress on our work of the CloudStack-KVM integration for Veeam B&R, and build excitement with the CloudStack & Veeam user community.

Thanks to great work in progress by @shwstppr @abh1sar @weizhouapache we've now a fully functional CloudStack-KVM integration that works with Veeam B&R v13. This work will be submitted to the community soon after a final round of checks, review and QA.

We're hopeful of targeting this for ACS 4.23, but there are some moving parts and dependency on the Veeam team to validate/certify CloudStack as a platform in Veeam.

The architecture looks like this, the integration is largely turnkey:

This will require some global settings to be enabled and configured, along with an option for a service account:

The integrations enables Veeam driven B&R with self-service backups coming in future version when B&R APIs are made available for UHAPI platforms:

Restore example:

Incremental backup example:

Restore of an incremental backup example:

PR: #12991

[mvds-vdsc]

Would this setup also work for Xenserver/XCP-NG Hypervisors? Since Veeam has an XCP-NG agent in Beta I hope your work would be reusable for XCP-NG. Great job by the way!

[yadvr]

I think Veeam already supports XCP-ng directly, though not keeping cloudstack aware of out of band changes (such as restoring backup such that they create new instances). That said, the agent based backups should work -- you should really check Veeam docs though.

[BaptisteTellier]

XCP-NG is supported by Veeam directly, it has nothing to do with UHAPI. Wait for official support in the coming month

[wido]

Looks great! Hope this motivates others to work on such integerations.

The first screenshot shows /ovirt-engine in the config. Does this integration share a lot of similarities on the Veem side with oVirt?

[shwstppr]

@wido yes, the integration API that Veeam supports via its UHAPI concept is largely the same as oVirt APIs
https://www.veeam.com/company/press-release/veeam-data-platform-v13-cyber-resilience-data-protection-security-ai.html

[yadvr]

@wido wot he said ^^ once our integration work is validated and tested by Veeam, they'll add a plugin/support for CloudStack as a platform; so longer term if everything goes well, users will see the management server end-points listed under CloudStack (and not ovirt -- this is just for development sake).

[notdrowningworking]

this will be very exciting

[flolicht]

That's awesome! This will make migrating existing VMware environments so much easier from the backup perspective.
Will the Veeam KVM implementation require specific primary storage backends, such as NFS or Ceph RBD?

[weizhouapache]

@flolicht
In Stage 1, the solution will support file-based storage, including NFS, local storage, and SharedMountPoint.
Both full backup and incremental backup will be supported, provided the images are in qcow2 format.

Support for Ceph and CLVM is planned for future stages, subject to feasibility.

[yadvr]

Wot ^ Wei said; we'll initially support NFS, local storage and shared mount point. Ceph and others will come eventually. (Backups of running VMs might work agnostic of storage type, based on how nbd/qemu works). Additionally, NAS B&R will also get several improvements, cross-zone DR support is already there and NAS B&R's incremental backup feature.

[BannerSmash]

Wot ^ Wei said; we'll initially support NFS, local storage and shared mount point. Ceph and others will come eventually. (Backups of running VMs might work agnostic of storage type, based on how nbd/qemu works). Additionally, NAS B&R will also get several improvements, cross-zone DR support is already there and NAS B&R's incremental backup feature.

Can you test this with iSCSI storage? Or possibly a restore from a storage snapshot? The main frustration is getting the data back into Cloudstack as a known resource.

[shwstppr]

@BannerSmash sorry but we are not considering iSCSI for the first iteration, but it is definitely something to look into in the future

[wido]

@BannerSmash sorry but we are not considering iSCSI for the first iteration, but it is definitely something to look into in the future

Understandable. I think that Ceph RBD would be even more useful as this is very widely used by CloudStack users.

[rodrigues8]

Amazing, I'm looking forward to the moment we can test this.

[shwstppr]

@rodrigues8 if you fancy, you may build and test the PR #12991

[yadvr]

@nvazquez @weizhouapache @shwstppr @abh1sar are now leading the project and best to advise further if anyone has any questions.