-
Notifications
You must be signed in to change notification settings - Fork 1.2k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
FRR 10.0.1 route maps filtering prefixes that should be sent #16367
Comments
Full running config of 172.18.0.5 (the receiver)
|
Running config for the instance running on the cluster |
logs of the instance running on the cluster |
Two identical match/set with different sequence numbers? Btw, could you please enable Also, I'm trying to understand the use case here with |
Ugh, I guess we can remove them
Yup, will do and re-run
The idea (maybe I am mis-using it) is to apply the local pref and then validate the ip later on to the following maps. I need it in case I want also to apply a community to the same prefixes |
@ton31337 here we go:
Show running config and full log attached : |
I checked only the first, seem to be there, so I'd say it's not related to the redundant match (although it makes sense to remove it). |
Is this happening during frr-reload, on boot, container restart? |
Definitely during reload. It's a battery of e2e tests, where on each test we run a different configuration and do some assertions. |
Could it be possible to have a full reload log? frr-reload.py --debug --reload /etc/frr/frr.conf. I'd like to understand what actually triggers this flake... |
Here's another run:
|
Last reload event: |
Sequence of reloads in the last 10 minutes Every reload sequence starts with Every cycle, we do a |
As discussed on slack, I attached the logs of the reloader and the generated configuration when running with 9.1 (the generated config is the same regardless). |
First, this is the wrong sequence of operations:
Looking at the logs you provided, it's not clear which AFI triggered this output (I assume IPv6):
Could you move Also, could you enable |
Sorry for the extreme delay @ton31337 , here I am with the additional information: Here we have From the logs the usual:
I applied the requested changes (define the prefix lists before the maps and adding log commands in the configuration): Attaching a tgz with:
Also, note that in a few reloads I am seeing
Not sure it's related or not. I am going to attach also a file with more reload events (reload_full.txt) which contains them too. |
Description
This is happening in metallb/frr-k8s#166 , where I am bumping the version used in frr-k8s from 9.1 to 10.0.1. The FRR-K8s CI is quite stable, while after the bump I am seeing a few flakes.
172.18.0.2 tries to announce 192.168.0.0/24 and 172.18.0.5 is not receiving it
I am going to attach the full configuration, but adding here the relevant snippets:
Neighbor route map:
Prefix lists:
Route maps:
Show bgp ipv4 on 172.18.0.2
Show bgp ipv4 on 172.18.0.5:
(as we can see some other instances were able to push the routes.
The logs say the prefix is being filtered by the route-map:
Version
How to reproduce
It's a flake, the same test sometimes passes and some other time it does not. I have the feeling that it is associated to prefixlist with community / local preferences. On the other hand, it happens often enough to be able to provide the right dump commands after a test fails.
Expected behavior
The neighbor should receive the prefixes
Actual behavior
The instance running on the cluster is not advertising the prefixes.
Additional context
No response
Checklist
The text was updated successfully, but these errors were encountered: