Search results

Thanks! Can you also post the syslog in the time period where the HA resource was unexpectedly started when the node was rebooted? The syslog should contain at least the task starts and ends of starting vm:118 and the pve-ha-lrm's output, ideally also the decisions from pve-ha-crm, which will be...

Hi! How does the crontab on the PVE node schedule the starting and stopping of the HA resource? If a HA resource is in the stopped state before the node is rebooted/shutdown and started, the HA resource should not start on its own again until it receives a request_start request, e.g. through...

Hm, yes these paths/files should exist, because these are the rrd databases, which store the data for the statistics shown in e.g. the summary graph and these should be created when these resources (node, storage or guests) are created. Is pvestatd running? Are there any other error messages in...

Hi! Could you also send the output for journalctl -u pve-ha-crm -u pve-ha-lrm --since '2026-01-31 22:25:00' --until '2026-01-31 22:30:00' on the involved nodes (especially the HA Manager)? This shows explicitly when the node maintenance mode is enabled/disabled and which migrations are done...

Hi! Which guide did you follow? Does the file /var/lib/rrdcached/db/pve-node-9.0/Lab exist? Is the name of the node itself Lab? Was the hostname changed inbetween? Is it possible that there is some IP address collision?

Hi! The HA groups are deprecated and will be removed eventually by Proxmox VE 10. The reason that the group API endpoints and parameters still exist is because of the state, where not all nodes have been upgraded yet and therefore only a part of the cluster can understand HA rules. As soon as...

Hi! A bad page with a non-zero reference counter usually means bad memory, a bad disk (if swap - e.g. hibernating - is used), or a kernel bug. Does the system use swap? Does this happen with other processes? In which kernel version did this kernel error start happening? Something like...

Hi! Memory hot-unplugging (as well as any other kind of hot-plugging/hot-unplugging) needs support from the guest operating system and it seems that it doesn't have that capability here. Does the VM have all the virtio-win drivers from the most recent virtio-win distribution installed? Is the...

Right, that could definitely be improved by using "Keep Together (positive)" and "Keep Separate (negative)" in the web interface or using only the "positive" and "negative" names there too. I'm not sure about including the rule name/description here, because the former isn't shown in the web...

From the old resources.cfg from Nov 12 it seems that the HA groups were never fully migrated. Was the ` at the end an artifact of embedding it as code int he forum or was that part of the file? Either way, great to hear that your problem has been solved!

I'm sorry, but I cannot reproduce this issue with the information you gave me. There's one newer version of pve-ha-manager on trixie/pve-test (5.1.0), but the bug fix in that version shouldn't be related to this here. Can you share the resources.cfg and rules.cfg for HA? If it's possible a...

On another note, what pve-ha-manager version is running on the nodes?

If the HA status report from above is the exact same as it was before setting pve1 in maintenance mode (the HA status report's timestamp is later than the syslog), then I cannot reproduce it with this configuration either. Are you sure that the rules are complete/not in conflict with other...

Thanks! If it's possible, it would be great to have a more complete reproducer for this to investigate the issue. The names can be changed, the only important part is that the changed names have the same alphabetical ordering (e.g. SN140 -> pve2, PVE003 -> pve1).

Hi! I have recreated your exact setup as described by the status output and rules config above and couldn't reproduce this either. What should happen is that as soon as pve1 is put in maintenance mode, the vm:102 will select a new node (it will be pve2 here as it's empty), vm:104 will select...

See my reply above, the syslog on the current HA Manager (master) node should show a regular error that the HA groups couldn't be migrated and the reason why. This should point you in the direction what is missing (e.g. maintenance mode, not quorate, non-existent nodes, etc.). All nodes must...

Hi! The versions of the node seem a bit out of date, pve-manager 9.0.5 is from mid-August. Could you try upgrading the nodes to see if this is fixed in a more current version? The HA groups should have been migrated with the upgrade from PVE 8 to PVE 9 automatically. The HA Manager tries this...

Was there any console output for the qmstart task for the VM 410? Is there anything shown at boot on the machine? Is the boot disk (scsi0) still intact (e.g., no fs corruption, system files are readable)? Else I could not see anything off from the syslog itself between start and end of the...

I cannot see anything obviously wrong about the config. Can you post the VM start log and the syslog around the VM startup?

From the error message it seems like that a random byte sequence was introduced... When you restart your machine / restart the pveproxy/pvedaemon services, then the Perl files are recompiled. I have never seen something similar before and can only guess that there must have been some corruption...