Github user zsxwing commented on the pull request:
https://github.com/apache/spark/pull/3527#issuecomment-65026993
> @zsxwing Note that the case you mentioned should no longer cause this
issue either, as we use an extra compressed data structure when dealing with
very large numbers of map partitions.
In extreme case, it's still possible. For example, assume that there are
10000 partitions in map side. If the user does not set a new `numPartition`,
there will be 10000 reducer. If all of these blocks are not 0, there will be
huge `MapStatus`s: 10000 * 10000 * 1 = 100MB. I'm not sure what the compression
ratio of `GZIPOutputStream` will be, but it may exceed `spark.akka.frameSize`.
Admittedly, this might be a user mistake and the user should set a proper
`numPartition`.
---
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]
