Closed famod closed 9 months ago
/cc @pedroigor (keycloak), @sberyozkin (keycloak)
Wondering if it could be related to https://github.com/quarkusio/quarkus/pull/37352 that landed in 3.7.0?
Linking to #38605 just in case too, and CC-ing to Martin @mkouba
Wondering if it could be related to #37352 that landed in 3.7.0?
Let me run a loop with 3.6.9...
Wondering if it could be related to #37352 that landed in 3.7.0?
Let me run a loop with 3.6.9...
I was unable to reproduce in 3.6.9, now trying with 3.7.3 plus git revert da170cebc42
...
Wondering if it could be related to #37352 that landed in 3.7.0?
Let me run a loop with 3.6.9...
I was unable to reproduce in 3.6.9, now trying with 3.7.3 plus
git revert da170cebc42
...
Yeah that's it - spot on @gsmet!
cc @xstefank
I was able to reproduce it, looking.
Thanks @xstefank! Crossing my fingers the fix might make it before 3.7.4, if it hasn't sailed already that is.
I wouldn't hold my breath as I will build it tomorrow morning (together with the 3.8.0 core artifacts). But it might land in 3.8.1 if not and 3.8 will be the continuation of the 3.7 branch (and a LTS) so it won't be a risky update.
@Ladicek should open a PR soon :)
Submitted #38888, which seems to fix this issue.
@famod if you have some cycles to test @Ladicek 's patch this afternoon, that would be nice.
Ah, I see you are already on it :).
We couldn't reproduce it anymore with the #38888
Describe the bug
Using
quarkus-keycloak-admin-client-reactive
andquarkus-smallrye-health
, we have a custom@Readiness
check in an@ApplicationScoped
bean that get's the request scoped instance injected. (In reality there is another@ApplicationScoped
in between, but that shouldn't matter.) In the respective readiness check method thatKeycloak
instance is used to get the realm and count its users.This works 99% of the time (and has worked 100% until a few weeks ago), but every now and then the check fails with:
or
As discussed in https://github.com/quarkusio/quarkus/discussions/38750, I changed this debug call to INFO and added the stacktrace which results in:
Curiously, only in a few of my many test sessions (session meaning running a readiness loop on a cleanly started server), I got this error instead:
So I went ahead and tried placing
@ActivateRequestContext
but failed to find the proper spot for this setup.Expected behavior
Consistent results with no error jitter.
Actual behavior
Failing check every now even though Keycloak is alive and kicking.
How to Reproduce?
no reproducer for now
Output of
uname -a
orver
No response
Output of
java -version
17.0.10
Quarkus version or git rev
3.7.3
Build tool (ie. output of
mvnw --version
orgradlew --version
)No response
Additional information
I ran my test sessions in a local docker compose setup but we also see it on our dev envs (also docker compose) and our Nomad envs - so I'd rule out network issues at this point.
See also discussion: https://github.com/quarkusio/quarkus/discussions/38750