[FLINK-30960] Fix a corner case oom error #77
Conversation
|
Thanks for opening this pull request! Please check out our contributing guidelines. (https://flink.apache.org/contributing/how-to-contribute.html) |
| * catch clause. In that case, we already have | ||
| * flushException cached, waiting for the next task | ||
| * manager thread's flush call which would throw a new | ||
| * FlushingRuntimeException causing job failure. |
There was a problem hiding this comment.
There is another case in this class where there is a flush() and is issues a RuntimeException, should we use the new Exception and catch it there as well; or can this case never loop. There are other calls to flush() - do we think these are not effected?
There was a problem hiding this comment.
I see, the close method calls flush as well. In this case, the scheduler is already shutdown before calling flush, so it will never loop.
It won't hurt if we throw the new FlushingRuntimeException here instead, as it will propagate up to be handled by TaskManager the same way as a regular RuntimeException.
|
is this the same as #5 |
I fixed a corner case out-of-memory error when we use a timed flushing interval, where the scheduler thread could repeatly throw a new RuntimeException wrapping the previous one as cause. This could lead to an infinite object reference chain.