Bugs in Create Cluster UI for IPv6 using LLA and for interface comments (v9.2.11)

[Threaded]

Goal: Creating a cluster using IPv6 interfaces for Ring0 and Ring1

Problems/bugs encountered:

1. Create Cluster UI won't allow selecting two different IPv6 interfaces using the same Link Local Address
2. Create Cluster UI only displays the address selected, but not the interface. At least for LLA it should be e.g. "fe80::11%nic1"
3. Create Cluster UI interface picker doesn't display comments for most interfaces, even though all have comments configured using 'Networks' UI
4. Create Cluster UI will only offer one IPv6 address for each interface, even when they have multiple

I worked around not supporting multiple addresses per interface by using vlans with one address each.

I haven't tested if I manually configuring the LLA addresses in corosync.conf works yet e.g. "fe80::11%nic1" and "fe80::11%vmbr0"
I assume it does and this is just a UI bug?


Dialog blocked because LLA s the same, even though the two Links are different interfaces:


Interface does not display comments for interfaces, even though comment added using Network dialog (see below)



Network dialog, showing configure comments:


/etc/network/interfaces file, generated by Proxmox Networks dialog

 

[longer]

For the comment issue, I think its the below bug:
https://bugzilla.proxmox.com/show_bug.cgi?id=6943

 

[shanreich]

There's also this issue for scoped addresses:
https://bugzilla.proxmox.com/show_bug.cgi?id=8014

 

[longer]

So, how can this error be fixed?

At the moment this isn't fixed yet. If you want, you can CC yourself on the bug report to stay updated on the progress.

 

[Threaded]

Thank you both very much for the bug references @longer @shanreich I have subscribed to them.

Wow - that IPv6 scoped addresses bug is a doozy! IPv6 support is fundamentally borked, right at the core address-parsing libraries used.

It gets worse, I tried to create a cluster with the LLA addresses and the experience was a disaster. It was not that it failed to add the 2nd node, but it left both the cluster node and and the added node unusable. The added node just stopped responding and wouldn't accept logins, and the cluster wouldn't do anything because it said no quorum. I tried adding the scope to the 'corosync.conf' file, but something keeps removing them. I feel like adding a node should be an atomic operation that should fail more gracefully; If the pre-add network tests fail, just don't proceed.

I'll have to declare failure on setting up a IPv6 cluster, or use ULA addresses I guess. It's a shame, because LLA addresses are exactly I the ideal approach for Corosync rings.