GitHub user brad-kaiser opened a pull request:
https://github.com/apache/spark/pull/19041
[SPARK-21097][CORE][WIP] Add option to recover cached data during dynamic
aâ¦
â¦llocation
## What changes were proposed in this pull request?
Add an option to to recover cached data when using dynamic allocation.
Currently when dynamic allocation decommissions an idle executor, any cached
RDD blocks on that executor will be lost. With this change those blocks will be
replicated to other executors. If there is not enough space for the cached
blocks, or any error is encountered, we will fall back to the original behavior
and drop the blocks.
Added the following config options
```
spark.dynamicAllocation.recoverCachedData
spark.dynamicAllocation.recoverCachedData.timeout
```
## How was this patch tested?
There are unit tests in RecoverCacheShutdownSuite and integration tests in
RecoverCachedDataSuite. I have also run benchmarks on a test cluster.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/brad-kaiser/spark recover-cached
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/spark/pull/19041.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #19041
----
----
---
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]
