You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
With @sergenyalcin, while doing some scale tests with 10K MRs with the official AWS provider and the no-fork external client implementation using the provider image index.docker.io/ulucinar/provider-aws-iam:v0.39.0-iamnofork, we've observed pod restarts due to a concurrent map iteration and writes to the same map:
We were able to see the writing site by running the provider with the Go race detector and setting the --max-reconcile-rate to 1000 and setting --poll to 10s:
What happened?
With @sergenyalcin, while doing some scale tests with 10K MRs with the official AWS provider and the no-fork external client implementation using the provider image
index.docker.io/ulucinar/provider-aws-iam:v0.39.0-iamnofork
, we've observed pod restarts due to a concurrent map iteration and writes to the same map:The relevant provider logs are as follows:
We were able to see the writing site by running the provider with the Go race detector and setting the
--max-reconcile-rate
to1000
and setting--poll
to10s
:Looks like the
resource.AnnotationChangedPredicate
's implementation is modifying the shared informer cache. We had better keep the cache items intact.How can we reproduce it?
What environment did it happen in?
Crossplane version: UXP
1.13.2-up.2
Official AWS provider:
index.docker.io/ulucinar/provider-aws-iam:v0.39.0-iamnofork
The text was updated successfully, but these errors were encountered: