Closed robert-scheck closed 3 months ago
No, your assertion is incorrect. keepalived states that because the route doesn't have an interface it cannot track the route. That is inevitably the case with a blackhole route since it cannot have an interface.
So, it is fully intended that in such a case a correct configuration always throws a warning in the logs? That's confusing for administrators from my point of view.
The way to stop the warning is to add no_track
to the route specification.
However, I have been thinking some more about this and I cannot see any reason why an interface is needed now in order to track the route. Accordingly commit a205e87 removes the requirement for a virtual route to have an interface specified for it to be tracked, and consequently no longer issues the warning.
Describe the bug
To Reproduce Well, it's a (general)
blackhole
route…to which interface shall this (general)blackhole
route be assigned? Aside of this, this wouldn't make sense to me. It looks in the configuration like this:Further on, the
blackhole
example inman keepalived.conf
also has no interface:Expected behavior Either the example in the man page should explain which interface to use with a (general)
blackhole
route (which still doesn't make sense to me), or the warnings for (general)blackhole
routes without interfaces should not be raised due to fixes in the code.Keepalived version
Distro (please complete the following information):
Details of any containerisation or hosted service (e.g. AWS) Real classic physical hardware.
Configuration file: (Please let me know if the example in the section "To Reproduce" isn't enough)
Notify and track scripts (The notify scripts are the default
primary-backup.sh
as provided in the examples/documentation)System Log entries
Did keepalived coredump? No
Additional context This behaviour did not exist with keepalived 1.3.5 as shipped with Red Hat Enterprise Linux 7 (yes, that was like a decade ago).