Github user JoshRosen commented on the pull request:
https://github.com/apache/spark/pull/5725#issuecomment-97145397
When running with off-heap, one concern is proper cleanup of managed memory
when tasks fail or are cancelled. Given that we currently do not share managed
memory between tasks running on the same executor, I'm thinking that we should
probably move MemoryManager from SparkEnv to TaskContext. This would let us
perform leak-detection when tasks complete; it also simplifies concurrency
issues (no contention for the MemoryManager's lock when allocating / freeing
blocks, provided that tasks are single-threaded) and reduces the chance that
we'll exhaust our page table limit, since each task will have its own
independent table.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
