fix(node): Local variables handle error #13827
Conversation
size-limit report 📦
|
|
hah. I was just testing the new implementation and ran into this exact problem, was about to open a gh issue. @timfish is it worth implementing some kind of heartbeat on the worker so we can detect (and log) when it's dead? |
|
And btw, a minimal reproducible example would likely be to throw an exception but immediately throw it away (+ let it be GC'd?) without doing a Sentry.captureException(). Perhaps that could be added as a test. I believe what's happening is exceptions that get caught and thrown away before the 1s releaseObject timer runs result in this lovely inspector error: |
|
As a final FYI i tested a try/catch in the worker and it has fixed the worker dying in my testing. |
If you have logging enabled via |
|
The fix for this has been released in v8.34.0! |
|
I just tried it out, weirdly objects of a class end up null :/ |
|
@chiragjn this looks unrelated to this issue so can you open a new issue with a reproduction? |
Ref #13768
Runtime.releaseObjectcan throw. I've reworked it to simplify and so that errors are caught.After a lot of testing, I've removed the memory leak test. I've tested this PR extensively on Node v20 + v22 and while memory usage does peak at ~350MB, it doesn't get above that after 15 minute of 1k errors per second. Memory usage varies from 150-300MB but I've concluded it's not actually leaking.