Is fence_pve abandoned by Proxmox?

[esi_y]

Debian still ships fence_pve [1] package [2], but Proxmox git has last activity dating back 9 years [3].

So the pacemaker fence agent got completely abandoned?

[1] https://manpages.debian.org/bookworm/fence-agents/fence_pve.8.en.html
[2] https://packages-picconi.debian.org/sid/fence-agents-pve
[3] https://github.com/proxmox/fence-agents-pve/tree/master/debian

 

[Neobin]

You seem to be the most bored person on this planet...

 

[esi_y]

This was asked a year ago here [1] with no further answer.

It's in the Proxmox git, so at least some time long ago, it had a use. The person got absolutely ignored [2] with his question in another thread.

At least in 2015/2016 there were some attempts to fix it [3ab].

Is everyone on those lists/threads bored or you just do not like my questions?

[1] https://forum.proxmox.com/threads/stonith-fencing-for-guest-vms.23193/#post-116284
[2] https://forum.proxmox.com/threads/pacemaker-manage-proxmox-vm.38942/#post-523616
[3a] https://lists.proxmox.com/pipermail/pve-devel/2015-August/016461.html
[3b] https://forum.proxmox.com/threads/right-permissions-to-stop-vm-with-fence_pve.26618/

 

[dcsapak]

mhmm there seems to be some overlap of package names

there is the debian package 'fence-agents': https://packages-picconi.debian.org/bookworm/fence-agents
which is the upstream https://github.com/ClusterLabs/fence-agents

in 'sid' it was split into multiple packages: see https://tracker.debian.org/pkg/fence-agents
which includes the 'fence-agents-pve' package that only contains one python script from upstream: https://github.com/ClusterLabs/fence-agents/blob/main/agents/pve/fence_pve.py
which is not from us (and never was)

our 'fence-agents-pve' https://git.proxmox.com/?p=fence-agents-pve.git;a=summary was an seemingly early attempt to repackage the 'fence-agents' package for our purposes but AFAIK this was never used by us really (and the naming is just coincidence)

hope that clears that up

 

[esi_y]

mhmm there seems to be some overlap of package names

I am not sure I got you here, it started with the upstream one:
https://git.proxmox.com/?p=fence-agents-pve.git;a=commit;h=24b72ae8cd6f0ebe9218645cd6223603f115136b

there is the debian package 'fence-agents': https://packages-picconi.debian.org/bookworm/fence-agents
which is the upstream https://github.com/ClusterLabs/fence-agents

in 'sid' it was split into multiple packages: see https://tracker.debian.org/pkg/fence-agents
which includes the 'fence-agents-pve' package that only contains one python script from upstream: https://github.com/ClusterLabs/fence-agents/blob/main/agents/pve/fence_pve.py

This is the case of each such agent.

which is not from us (and never was)

I never said it was. You were patching the upstream.

our 'fence-agents-pve' https://git.proxmox.com/?p=fence-agents-pve.git;a=summary was an seemingly early attempt to repackage the 'fence-agents' package for our purposes but AFAIK this was never used by us really

See e.g. https://lists.proxmox.com/pipermail/pve-devel/2015-August/016459.html

I suppose when you were patching it for upstream:
https://github.com/ClusterLabs/fence-agents/commit/c0a6b2567a6a04841adc9b991187cd51df19a635

... there was some use to it?

(and the naming is just coincidence)

Yeah this part I did not get what you were getting at.

EDIT: I think I got you now, yeah I do not mind how sid split it into individual agents, I was getting at the fact you were patching it for upstream once. How you called it in your repo or how sid now split it is not confusing me.

 

[dcsapak]

... there was some use to it?

i guess we wanted to use it and prepared the package, but never got around to it, at least i can't remember any time we would have used it (and we definitely don't use it at the moment, so it's just a remnant from an old approach probably)

I am not sure I got you here, it started with the upstream one:
https://git.proxmox.com/?p=fence-agents-pve.git;a=commit;h=24b72ae8cd6f0ebe9218645cd6223603f115136b

Yeah this part I did not get what you were getting at.

there is "our" fence-agent-pve (on git.proxmox.com) and the debian 'fence-agents' and 'fence-agents-pve' packages
"our" package is the repackage of 'fence-agents' in debian, not the 'fence-agents-pve' from debian (which is just a part of it)

I never said it was. You were patching the upstream.

yes, but you wrote:

Is fence_pve abandoned by Proxmox?

the 'fence_pve' tool was not from us and our patched package was not in use by us, so there is nothing to 'abandon' from our side here...

 

[esi_y]

there is "our" fence-agent-pve (on git.proxmox.com) and the debian 'fence-agents' and 'fence-agents-pve' packages
"our" package is the repackage of 'fence-agents' in debian, not the 'fence-agents-pve' from debian (which is just a part of it)

Noted, not an issue, neither did I think you were a maintainer.

yes, but you wrote: "Is fence_pve abandoned by Proxmox?"

My bad, what I meant was - did you stop "supporting" it?

the 'fence_pve' tool was not from us and our patched package was not in use by us, so there is nothing to 'abandon' from our side here...

I just can't quite believe you took it in 2011 and then went as far as patching it 2015, you even backport-fixed fencing_snmp.py in 2013, with no need for any of the users whatsover.

I understand it has nothing to do with the HA stack today, maybe it was never part of anything out-of-the-box, what I was after was ... this was all just for some random users (surely they had a use case)...

 

[dcsapak]

I understand it has nothing to do with the HA stack today, maybe it was never part of anything out-of-the-box, what I was after was ... this was all just for some random users (surely they had a use case)...

as i said AFAIU there was a plan to use them, but that never came to fruition. and looking at the git history on our side there was not much activity besides bumping and syncing with upstream (which is not that much work...)
sending patches upstream, even for things we don't use (especially if it concerns e.g. our api) is what we do sometimes, but that does not mean we support or maintain it...