Is it possible to have a longer cluster tasks-list available?

[scornelissen]

I'm looking for a solution to get longer task history in de Proxmox GUI. I found it it is the cluster tasks which is currently limited to 25 entries. I've been looking through the code, and it seems like is PVE::Cluster::get_tasklist() is returns always max 25 entries per node. My perl is not so good, so I could not really found why this limit exist.

As far as I can see, each node has a task list; the last entries of all nodes together show up in the cluster log. The length of tasks of the nodes can be specified on the api (/nodes/{node}/tasks, defaults to 50 entries). But the tasks list of the cluster seems to be limited by 25 entries per node.

 

[t.lamprecht]

Hi!

That specific list is streamed in real-time throughout the whole cluster, and effectively handled by pve-cluster (pmxcfs), so this is a hard limit there due to resource limitations from the cluster network.

We broadcast all currently running tasks (up to 32 KiB at max) plus 25 minus that number of running tasks of the last finished ones. While in theory that could be increased it should not be required, you can poll each node for much more tasks without hitting the more bandwidth constraint cluster network.
But just for the record and completness sake, that maximal number of last inactive tasks would need to be updated in the active_workers method of the RESTEnvironment base module (shipped by libpve-common-perl package).

What would be your use case here? Maybe we can help out with some possible workarounds...

 

[scornelissen]

I had an user which would look further back in history from the Proxmox UI. If something in history happened (and we didn't know what VM) we could not found it back in the cluster log. Now I know it's a hardcoded limit (and better be not increased) I think I will look for some way to store the tasks elsewhere.

It's limited because of the max limit 30MB of RAM as described here: https://pve.proxmox.com/wiki/Proxmox_Cluster_File_System_(pmxcfs)?

 

[BloodyIron]

I'm necro-ing this thread because I actually have a functional need for a more consistent Cluster Task list than we are seeing lately.

Client's env is v8.1.x Proxmox VE Cluster, and when I review the last week of backup tasks a lot of the tasks are not being shown in the Cluster Task list even a week ago. I know the backups happened, but I rely on this central list to be consistent and provide an overview of the backup history (errors/warnings/none). This would be a bigger problem with a larger cluster.

I can appreciate there is a balance to strike for performance Cluster Task lists, but this is clear-text... surely background tasks can maintain a consistent list as time goes on so said fidelity isn't lost. Isn't that point of corosync, a clustered filesystem that said info could be kept on so all nodes can have consistent info? Also, why isn't the Cluster Task list an available section at the "Cluster" level? I really would like a section in there where I can filter the results... go back further in time... etc...

Any chance we can have this topic revisited? I see there being tangible value in the Cluster Task aspect being given love (whenever is another matter of course).