Added by Viktor Gurov over 2 years ago. Updated almost 2 years ago.
100%
Description
If IPv6 link-local address is used as `rainterface`, the status of the CARP VIP is not checked (get_carp_interfaces_status())
related to #11103
Related issues
Updated by Viktor Gurov over 2 years ago
Updated by Viktor Gurov over 2 years ago
Updated by znerol znerol over 2 years ago
It is in fact expected behavior that radvd is starting on both hosts. radvd is supposed to send RAs for a given interface iff the link-local address configured in AdvRASrcAddress is present on the interface.
This works well in Linux because Keepalived makes sure that the virtual link-local address is configured on one node only. In BSD/CARP however, a virtual CARP address is always configured on all nodes. It is just flaged with MASTER/BACKUP.
Regrettably radvd is not aware of the address flags in BSD. Thus, it will send RAs regardless of whether a link-local address is flagged with MASTER or BACKUP. I intended to report that upstream, but honestly forgot about it.
Thus, I agree that we need a workaround in pfSense until upstream radvd is patched to respect MASTER/BACKUP address flags on BSDs.
Updated by znerol znerol over 2 years ago
Upstream issue: https://github.com/radvd-project/radvd/issues/162
Updated by Jim Pingle over 2 years ago
Updated by Viktor Gurov about 2 years ago
Applied in changeset abc7b3056fafb57e8941103f2565b5b113edd177.
Updated by Jim Pingle about 2 years ago
Updated by Jim Pingle about 2 years ago
Updating subject for release notes.
Updated by Jim Pingle almost 2 years ago
Seems to be OK. With radvd set to use an LL VIP I still only see radvd running on the node with master status on its VIPs.