Open purajit opened 1 year ago
Are you passing any flags on the argocd app diff/manifests
command, or just the app name?
I have seen the same when I upgraded from argo 2.7.2 to argo 2.8.2, setting the redis.compression
config to none seems to fix this.
Also confirmed that adding redis.compression: none
in ConfigMap argocd-cmd-params-cm
solves the issue. I'm also using ArgoCD v2.8.4.
Are you passing any flags on the
argocd app diff/manifests
command, or just the app name?
Hey apologies, really need to figure out my GH notifications game.
Not passing in any other flags to the argocd
commands. I don't have a cluster without compression available right now, but I believe I had this issue with and without --local
as well.
same issue in 2.10.3. redis.compression: none as workaround solves
any idea? @aborilov
Might be related to https://github.com/argoproj/argo-cd/issues/5068
Checklist:
argocd version
.Describe the bug
ArgoCD 2.8 enables Redis compression by default. However, after upgrading, I found I was unable to run commands like
argocd app diff/manifest
, since it would fail withEverything else runs without issue - syncing, UI browsing, etc.
To reproduce Unsure what to mention here.
What I've tried that didn't help:
REDIS_COMPRESSION
isgzip
on all podsredis.compression
config togzip
--redis-compress=gzip
as a server argumentWhat works:
redis.compression
config tonone
Current configuration:
--insecure --enable-gzip
(we handle TLS termination ourselves)public.ecr.aws/docker/library/haproxy,redis
repositories, with HA enabled.Expected behavior
Everything should "just work", and key formation should be consistent.
Version
We're using the ArgoCD helm chart (v5.46.7).
Logs
I cleared out some cache keys, and then ran
redis-cli monitor | grep <application>
on the Redis pods:Note the lack of the suffix on the last one.
I could potentially add a ton of logging, build argo-cd and create an image, and deploy that to get more information, but I was hoping to see if anyone has dealt with this or knows what might be happening first. Looking at the code, I can't see why the key formation would ever be inconsistent.
Some relevant code paths: