for future reference here is the xl dmesg at this point xldmesg.txt

Thirty to forty seconds of clock drift after a power outage, and then the backup goes through once you fix the time. That could be coincidence, but it might not be. I don't know enough about how the XAPI client handles time skew to say whether it would surface as a BodyTimeoutError rather than an auth or TLS failure, so I could easily be wrong here. If it fails again, could you look at what the host clock is doing before you correct it? @pierrebrunet from the XO team is already on this one, and a second data point either way would tell him whether it's worth pulling on. Thanks for putting the logs on the pastebin, much easier to read that way.

Indeed, it should be detected/reported correctly to avoid losing time finding the problem. Let me ping @julienXOVates

@pkgw Planning an RPU as a whole is a nice idea in theory, but it's inherently designed to co-exist with normal pool management operations (you should still be able to migrate VMs between hosts B and C during it, or start new ones), we are not "pre-booking" memory and other resources for groups of VMs, otherwise it can quickly make the rest of the pool hard to manage while some VM migrations could still be far away. Regarding the narrower case - I'll note this down and see if we can improve things, thank you!

@itservices Hello, I'm taking this issue from Pierre. Did you perform a full restart of the XCP-ng host since the issue occured ? Or just restarted the toolstask ?

The 2026-05-28 build still being good narrows the window a lot more than "sometime since November". What I keep coming back to is the shape of the failure: 14 out of 15, then 5 out of 6, so it is always exactly one that falls over and never the whole run. I don't know whether that points at a per-VM timeout or at something the last task in a run does differently, and someone on the XO team will read that better than me. @pierrebrunet still needs /var/log/xensource.log from the pool master covering one failed run's window, so even a slice from a job where only one VM failed should be enough.