bandwidth: handle nil bandwidth in CHECK#1222
Conversation
Somehow we missed this case; if CHECK is called with no bandwidth configuration, we segfault. Oops. Fixes: containernetworking#1221 Signed-off-by: Casey Callendrello <c1@caseyc.net>
|
Thanks for the quick fix - is there an ETA on a new release that might include this? I know v1.9.0 JUST came out. |
|
@brandond releases are cheap. I'll take care of it. |
|
@brandond any chance you could verify in your environment that this fixes the issue before we cut a release? |
|
This was reported by a k3s community member, using their own cri-o runtime instead of the containerd service that comes with k3s... so already outside something we would normally support. @adelton you should be able to overwrite the k3s bandwidth plugin binary with one built directly from this project to test - if you are able. |
|
What is the canonical way to build from master, ideally in some containerized fashion so that the environment is the same as the one used to build binaries for releases? |
|
@adelton it's not built in a container, and regardless it shouldn't really matter as this is pure-go and the NPE isn't dependent on go version or anything else. Just run the build script: plugins/.github/workflows/release.yaml Lines 24 to 28 in 8c4a6b3 |
|
I confirm that on the setup described in k3s-io/k3s#13344 (comment), when I run the stream of SIGSEGV errors in journal stops and the pods all get to either Completed or Running. So the code change causes a positive change in behaviour, and I dare to say, fixes the issue. |
|
Fantastic. @squeed could we get v1.9.1? |
|
ping @squeed - when can we expect to see a new release? |
Somehow we missed this case; if CHECK is called with no bandwidth configuration, we segfault. Oops.
Fixes: #1221