PVE 6 support

[trickert76]

Proxmox v6 Support - in my environment it works on a fresh Debian Buster.

[AleRousse]

Unfortunately i can't achieve to make this works with a debian 10 machine.
I am getting error when the role is trying to setup cluster with pvecm create.

I found that in proxmox VE 6.X you need to specify --linkX and not -bindnetX_addr and ringX_addr
You can verify this by the following:
https://forum.proxmox.com/threads/separate-cluster-network-wiki-bindnet0_addr-command.25342/
https://pve.proxmox.com/pve-docs/pve-admin-guide.html#_cluster_network

Let me know what you think about it

EDIT: Need to specify that the role seems to fails only when you try to configure it in a cluster as pvecm will only be called in that state

[lae]

Posts 5 and 6 are the relevant posts in the linked thread (post 5 permalink: https://forum.proxmox.com/threads/separate-cluster-network-wiki-bindnet0_addr-command.25342/#post-259509)

Reading through the above, and given the fact bindnet* is gone entirely, I think duplicating the initialization task (

ansible-role-proxmox/tasks/pve_cluster_config.yml

Line 37 in 147d366

- name: Initialize a Proxmox cluster

) using the link* parameters, templated via a new role variable of the same name, with a when condition (distribution == buster) might be more maintainable. The README would then just have a note saying the bindnet/ringnet parameters are for PVE 5.x.

Thoughts?

[AleRousse]

@lae Yeah I think to duplicating and making task distribution specific would be more clear.

Also I am experiencing another task failing in the following:

ansible-role-proxmox/tasks/pve_add_node.yml

Line 9 in 147d366

-ring0_addr {{ pve_cluster_ring0_addr }}{% if pve_cluster_ring1_addr is defined %} \

Again I think the solution would be duplicating the task and making it distribution specific

[lae]

Oh, I forgot to do a quick search for ring addr too.

@trickert76 what do you think? if you're fine with it, could you make the changes, too?

[lae]

The conditional here can be removed entirely - I missed this when removing support for 4.x.

[trickert76]

Great - I'll change that too and make a new PR. I would make two tasks of it and add a when condition

[lae]

@trickert76 that change should be in this PR, not a separate one (it's a required change for the role to continue working in clusters for PVE 6 as far as I can tell?)

[trickert76]

I don't understand that error in Travis.

[trickert76]

How could I define the distribution version in the test - I only see in tests/install.yml a profile - but it should be a second test for Buster too. Also I don't understand - if it is a Stretch test, why should it be changed, when the "when" condition is working.

[lae]

@trickert76 this is a tricky one - the space before -use_ssh is missing now so it tries to add {{ host_ip }}-use_ssh as a node. I'm not sure if this is what causes timeouts for the first two, but it's visible in the last one. Try moving -use_ssh to the line before.

As for adding tests for PVE 6, the following steps are needed:

• add 3 new nodes to tests/inventory
• update tests/install.yml by adding another item to test_profiles, profile = debian-buster and prefix = proxmox-6-

But right now there isn't a debian-buster test profile in the test suite yet, so it won't work. Let me get to updating the test suite and I'll add a commit for PVE 6 tests.

[lae]

@AleRousse Could I have you try to deploy this a PVE 6 cluster with this branch and report back if you have any other issues?

[AleRousse]

@trickert76 and @lae I confirm that it works well for a PVE6's cluster
Tested with the following feature:

• PVE 6 Base installation
• Storage content with LVM
• Clustering 2 nodes with Corosync 3.0.2

Thanks all!