There are 2 updates below. tldr;
- Update 1: Remove ibev-perl and restart pvescheduler
- Update 2: Restarting pvescheduler may not be enough. Instead stop pvescheduler and then start it again.
Original Problem:
We have two clusters, one has 3 nodes and the other has 2. Both clusters are kept up to date, use PBS and had working scheduled backups before the upgrade to PVE 7.4.
After upgrading to PVE 7.4 the 2 node cluster is working (and backing up) fine. On the 3 node cluster only guests from one of the nodes are being backed up. There are no errors - anywhere. There is nothing about trying to backup the guests from the other two nodes. They seem to just be ignored.
I have attached some files:
- Pve03.txt has the file versions of the PVE nodes.
- All 3 nodes have the same versions except for the old kernel versions (they are all running the same kernel.)
- Pbs.txt has the file versions of the backup server. The backup server is running as a VM on the node that the backups are working on.
- jobs.cfg.txt are the backup job files from /etc/pve - as one would expect the file contents match.
- email-log.txt is the text of a scheduled backup email that shows all of the guests that were backed up and those that should have been backed up. No errors are shown in this log either.
- vzdump.cron only has a PATH statement:
PATH="/usr/sbin:/usr/bin:/sbin:/bin"
If I create a new duplicate backup schedule only the guests on Pve03 are backed up.
- If I manually run the backup schedule all selected guests on all 3 nodes are backed up.
If I create separate backups for each node then the two problem nodes backups do "not" run on schedule. No errors, no logs. Just ignored. (They are in jobs.cfg.txt)
- They will run if started manually.
If it matters the two nodes where the backups do no run were created first. They were added to a 2 node cluster. They were running PVE 5x. Then the third node was added at a later date under PVE 6x (I think.) The third node is the only node that scheduled backups run on.
Thanks for a great set of products!
Update 1:
Using what nevakee posted below I removed `libev-perl` from my 2 problem nodes and restarted pvescheduler. Because I have "Repeat Missed" checked in my backup settings the guests on each of the nodes started backing up immediately.
- Both nodes have been running PVE since at least 5.x
- I verified that the other "newer" nodes did not have `libev-perl` installed.
Note: Updating this post caused the while thread to be embargoed until it was approved. That's why it went missing for a period of time.
Update 2 - March 30:
After some updates to PVE the newest node stopped performing scheduled backups. This is the node that always worked.
- Restarting pvescheduler did not fix the problem.
- After the second round of updates and an update to Proxmox backup the problem persisted.
It was found (at least on this node) that restarting pvescheduler won't work. The pvescheduler service has to be stopped and then started. This fixes its problem.
- Update 1: Remove ibev-perl and restart pvescheduler
- Update 2: Restarting pvescheduler may not be enough. Instead stop pvescheduler and then start it again.
Original Problem:
We have two clusters, one has 3 nodes and the other has 2. Both clusters are kept up to date, use PBS and had working scheduled backups before the upgrade to PVE 7.4.
After upgrading to PVE 7.4 the 2 node cluster is working (and backing up) fine. On the 3 node cluster only guests from one of the nodes are being backed up. There are no errors - anywhere. There is nothing about trying to backup the guests from the other two nodes. They seem to just be ignored.
I have attached some files:
- Pve03.txt has the file versions of the PVE nodes.
- All 3 nodes have the same versions except for the old kernel versions (they are all running the same kernel.)
- Pbs.txt has the file versions of the backup server. The backup server is running as a VM on the node that the backups are working on.
- jobs.cfg.txt are the backup job files from /etc/pve - as one would expect the file contents match.
- email-log.txt is the text of a scheduled backup email that shows all of the guests that were backed up and those that should have been backed up. No errors are shown in this log either.
- vzdump.cron only has a PATH statement:
PATH="/usr/sbin:/usr/bin:/sbin:/bin"
If I create a new duplicate backup schedule only the guests on Pve03 are backed up.
- If I manually run the backup schedule all selected guests on all 3 nodes are backed up.
If I create separate backups for each node then the two problem nodes backups do "not" run on schedule. No errors, no logs. Just ignored. (They are in jobs.cfg.txt)
- They will run if started manually.
If it matters the two nodes where the backups do no run were created first. They were added to a 2 node cluster. They were running PVE 5x. Then the third node was added at a later date under PVE 6x (I think.) The third node is the only node that scheduled backups run on.
Thanks for a great set of products!
Update 1:
Using what nevakee posted below I removed `libev-perl` from my 2 problem nodes and restarted pvescheduler. Because I have "Repeat Missed" checked in my backup settings the guests on each of the nodes started backing up immediately.
- Both nodes have been running PVE since at least 5.x
- I verified that the other "newer" nodes did not have `libev-perl` installed.
Note: Updating this post caused the while thread to be embargoed until it was approved. That's why it went missing for a period of time.
Update 2 - March 30:
After some updates to PVE the newest node stopped performing scheduled backups. This is the node that always worked.
- Restarting pvescheduler did not fix the problem.
- After the second round of updates and an update to Proxmox backup the problem persisted.
It was found (at least on this node) that restarting pvescheduler won't work. The pvescheduler service has to be stopped and then started. This fixes its problem.
Last edited:
