Github user vanzin commented on a diff in the pull request:
https://github.com/apache/spark/pull/6648#discussion_r41546189
--- Diff:
network/shuffle/src/main/java/org/apache/spark/network/shuffle/ExternalShuffleBlockResolver.java
---
@@ -163,12 +163,12 @@ public void registerExecutor(
/**
* Obtains a FileSegmentManagedBuffer from a shuffle block id. We expect
the blockId has the
- * format "shuffle_ShuffleId_MapId_ReduceId" (from ShuffleBlockId), and
additionally make
- * assumptions about how the hash and sort based shuffles store their
data.
+ * format "shuffle_ShuffleId_MapId_ReduceId_StageAttemptId" (from
ShuffleBlockId), and
+ * additionally make assumptions about how the hash and sort based
shuffles store their data.
*/
public ManagedBuffer getBlockData(String appId, String execId, String
blockId) {
String[] blockIdParts = blockId.split("_");
- if (blockIdParts.length < 4) {
+ if (blockIdParts.length != 5) {
--- End diff --
Hmm... is the contract that a new shuffle service should work with apps
running an old Spark version? In that case, shouldn't this have different code
paths for `length == 4` and `length == 5`?
---
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]
