holdenk commented on a change in pull request #29211:
URL: https://github.com/apache/spark/pull/29211#discussion_r465926745
##########
File path:
core/src/test/scala/org/apache/spark/storage/BlockManagerDecommissionIntegrationSuite.scala
##########
@@ -266,18 +266,17 @@ class BlockManagerDecommissionIntegrationSuite extends
SparkFunSuite with LocalS
val execIdToBlocksMapping = storageStatus.map(
status => (status.blockManagerId.executorId, status.blocks)).toMap
// No cached blocks should be present on executor which was decommissioned
-
assert(execIdToBlocksMapping(execToDecommission).keys.filter(_.isRDD).toSeq ===
Seq(),
+ assert(
+ !execIdToBlocksMapping.contains(execToDecommission) ||
+ execIdToBlocksMapping(execToDecommission).keys.filter(_.isRDD).toSeq ===
Seq(),
"Cache blocks should be migrated")
if (persist) {
// There should still be all the RDD blocks cached
assert(execIdToBlocksMapping.values.flatMap(_.keys).count(_.isRDD) ===
numParts)
}
- // Make the executor we decommissioned exit
- sched.client.killExecutors(List(execToDecommission))
-
- // Wait for the executor to be removed
- executorRemovedSem.acquire(1)
+ // Wait for the executor to be removed automatically after migration.
+ assert(executorRemovedSem.tryAcquire(1, 5L, TimeUnit.MINUTES))
Review comment:
It doesn't take up to 5 minutes in my experience, I'm just picking a
high enough timeout it won't flake. I'm not open to changing this, I've dealt
with too many flaky tests from timeouts that were set too low.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
