[GitHub] [spark] HeartSaVioR commented on a diff in pull request #42895: [SPARK-45138][SS] Define a new error class and apply it when checkpointing state to DFS fails

["HeartSaVioR (via GitHub)" <gi...@apache.org>]

HeartSaVioR commented on code in PR #42895:
URL: https://github.com/apache/spark/pull/42895#discussion_r1326687236


##########
common/utils/src/main/resources/error/error-classes.json:
##########
@@ -314,6 +314,18 @@
       "<details>"
     ]
   },
+  "CANNOT_WRITE_STATE_FILE" : {

Review Comment:
   How about considering symmetric name? For the base error class of failure on loading state, we use CANNOT_LOAD_STATE_STORE. Probably better to sync between two.



##########
sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/state/HDFSBackedStateStoreProvider.scala:
##########
@@ -135,17 +135,15 @@ private[sql] class HDFSBackedStateStoreProvider extends StateStoreProvider with
 
     /** Commit all the updates that have been made to the store, and return the new version. */
     override def commit(): Long = {
-      verify(state == UPDATING, "Cannot commit after already committed or aborted")
-
       try {
+        verify(state == UPDATING, "Cannot commit after already committed or aborted")
         commitUpdates(newVersion, mapToUpdate, compressedStream)
         state = COMMITTED
         logInfo(s"Committed version $newVersion for $this to file $finalDeltaFile")
         newVersion
       } catch {
-        case NonFatal(e) =>
-          throw new IllegalStateException(
-            s"Error committing version $newVersion into $this", e)

Review Comment:
   Shall we check what is logged for `this`? Does it include more information about state store metadata like operator ID, partition ID, storeName, etc? If then might be better to keep it.



-- 
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.

To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org