Memory leak when requesting new token

[checketts]

I'm seeing memory leaks (heap growing until application crashes). The timing aligns with when my token refreshes are happening. And when doing a heap dump, it seems to be pointing to AbstractUaaTokenProvider:requestToken.

Here is the heap info:

Is clean up or a timeout required for the request?

[checketts]

My current theory is that the unlimited retries here: https://github.com/cloudfoundry/cf-java-client/blob/master/cloudfoundry-client-reactor/src/main/java/org/cloudfoundry/reactor/util/Operator.java#L209 might be related. I've adding some debugging and limiting it to 100 retries to see if I can get more insight into the behavior

[valentinEmpy]

Just as a note, the infinite retries have been a target of the following PR:
#1029
We experienced sporadic IPBased rate limit exceeded due to it and also suspected Netty connection/thread leaks, however we have not confirmed it.

[checketts]

Thanks for the pointer @valentinEmpy. I'll be able to confirm the connection/thread leaks. I hard coded in a 100 time retry just to be absurd, but not infinite in an equivalent fix.

I have 3 instances that are running my code and predictably 2 of the 3 run into this leak when the token expires (20 hours in my case). Interestingly the 2 that have trouble are calling CF APIs that are remote, while the 1 that doesn't have trouble is calling the API it is hosted in.

[checketts]

Should I close this now that #1029 is merged in? Or wait for a release?

[twoseat]

If you're able to use the build snapshot to satisfy yourself that this handles your scenario that would be great.

[twoseat]

I just spotted that you've done that already! If you're happy you can close this.