Github user ksakellis commented on a diff in the pull request:
https://github.com/apache/spark/pull/4909#discussion_r25964901
--- Diff: core/src/main/scala/org/apache/spark/deploy/master/Master.scala
---
@@ -467,7 +467,9 @@ private[spark] class Master(
* two executors on the same worker).
*/
def canUse(app: ApplicationInfo, worker: WorkerInfo): Boolean = {
- worker.memoryFree >= app.desc.memoryPerSlave &&
!worker.hasExecutor(app)
+ worker.memoryFree >= app.desc.memoryPerSlave &&
+ !worker.hasExecutor(app) &&
+ !app.removedExecutors.exists(_.worker == worker)
--- End diff --
I don't think this is a good idea as is. You can have an executor that died
because of transient failures(network blips) and if you blacklist them, they'll
never be able to come back. This means that your cluster will monotonically
decrease in size even if the once bad executors can in theory be relaunched
(unless you add more hosts dynamically). If you want to do this, i'd recommend
having a TTL on the blacklist (has been removed since a specific time).
---
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]
