kaniko hangs for minutes doing nothing when destination is insecure http registry #1933
Labels
area/performance
issues related to kaniko performance enhancement
area/registry
For all bugs having to do with pushing/pulling into registries
kind/bug
Something isn't working
priority/p2
High impact feature/bug. Will get a lot of users happy
Actual behavior
Kaniko hangs for minutes twice:
Expected behavior
Kaniko should not try HTTPS without timeout when given
--insecure-registry
parameter, such that it does not hang for minutes doing nothing.To Reproduce
Reproduced on Digital Ocean and Linode managed kubernetes.
Steps to reproduce the behavior:
kubectl run kaniko --restart=Never --image=gcr.io/kaniko-project/executor:debug -- \ --log-format=text \ --verbosity=trace \ --log-timestamp \ --context=git://github.com/arve0/kaniko-repro \ --insecure-registry=registry.default \ --destination=registry.default/arve0/kaniko-repro \ --force # force is required on linode, as runc detection does not work correctly
Alternative Reproduction
Installs registry and runs reproduction:
git clone https://github.com/arve0/kaniko-repro cd kaniko-repro ./reproduce.sh
Additional Information
Dockerfile, seem to be any Dockerfile. Has reproduced with
Build Context, no files needed. Reproduced with both mounted volume and git context. https://github.com/arve0/kaniko-repro
Kaniko Image (fully qualified with digest), both
:latest
and:debug
. Debug digest:sha256:fcccd2ab9f3892e33fc7f2e950c8e4fc665e7a4c66f6a9d70b300d7a2103592f
.Trace logs:
kill -ABRT 1
on first hang: https://github.com/arve0/kaniko-repro/blob/master/linode-start-kill.logkill -ABRT 1
on second hang: https://raw.githubusercontent.com/arve0/kaniko-repro/master/linode-push-kill.logRunning goroutines when hanging
What have I tried?
I've search issues after "hang", and found similar behavior described in #1287. Workaround in comment does not work. Also, my cluster have internet access, so not same issue I believe.
Adding HTTPS on port 443 to registry resolves the issue.
Triage Notes for the Maintainers
--cache
flagThe text was updated successfully, but these errors were encountered: