XCP-ng 8.3 updates announcements and testing

[acebmxer]

@dthenot

I have been able to migrate all vms over to qcow2. Think shutting down the vms and booting backup. Alos if anything from this thread might have had an impact. https://xcp-ng.org/forum/topic/12087/backups-with-qcow2-enabled/9

[dthenot]

@Andrew Hello,

I have been able to find the problem and make a fix, it's in the process of being packaged.
I can confirm it only happen for file based SR when using purge snapshots.
For some reason, the vdi type of CBT_metadata is cbtlog for FileSR but stays the image format it was for LVMSR
And it would make a condition fail during the list_changed_blocks call.

[olivierlambert]

Nice catch @dthenot !

[Andrew]

@dthenot Great! I'm happy I was able to help test it. I look forward to the update release.

Interesting note, CR is faster when the snapshots are not deleted.... or CR is faster because of the update, I'll test again after the fix.

[rzr]

Feature fixes, security and maintenance update candidates for you to test!

This release batch contains fixes on the major storage feature previously announced,
read the RC2 announcement for QCOW2 image format support for 2TiB+ images.

The whole platform has been hardened with back-porting security patches from the latest version of OpenSSH.

An additional driver fix is part of this minor package set.

What changed

Storage

QCOW2 image format support is the major feature of this release batch,
check related announcement in forum.

Some fixes have been applied to fix issues found during the testing phase. Many thanks go to @Andrew who found a CBT-related bug on file-based SRs!

• sm: 3.2.12-17.5
  • Fix a regression on CBT (Changed block tracking) on file-based SRs (EXT, NFS, ...), causing backup jobs using the "purge snapshot data when using CBT" option to create full backups each time instead of deltas.
  • Deactivate unused LVM snapshot base before deletion to prevent LVM leak. This fix is not related to the QCOW2 feature, but is important and localized enough for us to provide it in addition the other changes.
  • Minor fix that prevents a warning when updating the package.
• blktap: 3.55.5-6.5
  • Fix install warning when triggering mdadm to generate a udev rule.

Network

• openssh: Update to 9.8p1-1.2.3
  • Two vulnerabilities disclosed along with the OpenSSH 10.3 release have been fixed.
    • In authorized_keys, when principals="" was defined along with a CA with a common CA, an interpretation error occurred, which could lead to unauthorized access.
    • When one ECDSA algorithm was active, it activated all others regardless of their configuration. (By default, all ECDSA algorithms are active.)
  • For more details please track the upcoming Vates Security Advisories.

Drivers updates

More information about drivers and current versions is maintained on the drivers wiki page.

• qlogic-fastlinq-alt: 8.74.6.0-1
  • Fixes 2 issues in the qede module driver:
    • Driver does not retain configured MAC and MTU post reset recovery
    • Driver does not recover from TX timeout error

Versions:

• blktap: 3.55.5-6.4.xcpng8.3 -> 3.55.5-6.5.xcpng8.3
• openssh: 9.8p1-1.2.2.xcpng8.3 -> 9.8p1-1.2.3.xcpng8.3
• sm: 3.2.12-17.2.xcpng8.3 -> 3.2.12-17.5.xcpng8.3

Optional packages:

• qlogic-fastlinq-alt: 8.70.12.0-1.xcpng8.3 -> 8.74.6.0-1.xcpng8.3

Test on XCP-ng 8.3

If you are using XOSTOR, please refer to our documentation for the update method.

yum clean metadata --enablerepo=xcp-ng-testing,xcp-ng-candidates
yum update --enablerepo=xcp-ng-testing,xcp-ng-candidates
reboot

The usual update rules apply: pool coordinator first, etc.

What to test

The most important change is related to storage: adding QCOW2 support also affects the codebase managing VHD disks. What matters here is, above all, to detect any regression on VHD support (we tested it deeply, but on this matter there's no such thing as too much testing). Of course, you are also welcome to test the QCOW2 image format support.

See the dedicated thread for more information.

Other significant changes requiring attention:

• SSH connectivity

And, as usual, normal use and anything else you want to test.

Test window before official release of the updates

~4 days

We would like to thank users who reported feedback on the QCOW RC2 release: @acebmxer, @andrew, @bufanda, @flakpyro, @jeffberntsen, @ph7

[acebmxer]

@rzr

installed updates will report back.

Update - I had migrated vms back over to vhd prior to update release. I have migrated 2 vms back over to qcow2 and the initial backup ran successfull. Ran a second delta backup and that as well was successful with out issues. Backups happen very quickly now. But it appears the % and progress bar are working.

When CBT is enabled on the vm vdi. They show up as needing to be coalesced. VMs without CBT enabled the vdis are coalesced.

Will continue to monitor.

Once the coalesence hits 2 for the vm. The vm is skipped form future backups until cleared. (shutting down the vm will allow the coalescence to happen.
2026-04-23T19_52_34.694Z - backup NG.txt

[Andrew]

@rzr XCP 8.3 pools updated and running.

CR delta backup snapshot problem corrected and working now.

SSH from old system to XCP displays the warning (per documentation).

[acebmxer]

@rzr

I thought i pressed save after editing the backup job to enable the purge snapshot when using cbt.

After re-enabling and clicking save. All is good now. No errors no stuck coalescence after backups.

[rzr]

This post is deleted!

[MajorP93]

I updated my test environment and performed a few tests:

• migrating VMs back and forth between VHD based NFS SR and QCOW2 based iSCSI SR --> VMs got converted between VHD and QCOW2 just fine, live migration worked
• creation of rather big QCOW2 based VMs (2.5+ TB)
• NBD-enabled delta backups of a mixed set of VMs (small, big, QCOW2, VHD)

All tests worked fine so far. Only thing that I noticed: When converting VHD-based VMs to QCOW2 format I was not able to storage migrate more than 2 VMs at a time. XO said something about "not enough memory". That might be related to my dom0 in test environment only having 4GB of RAM. Maybe not related to VHD to QCOW2 migration path. I never saw this error in my live environment where all node's dom0 have 8GB RAM.

Update candidate looks good so far from my point of view.