gitlw commented on code in PR #12029:
URL: https://github.com/apache/kafka/pull/12029#discussion_r850657561
##########
core/src/test/scala/unit/kafka/server/ReplicaManagerTest.scala:
##########
@@ -2615,7 +2615,10 @@ class ReplicaManagerTest {
@Test
def
testStopReplicaWithDeletePartitionAndExistingPartitionAndNewerLeaderEpochAndIOException():
Unit = {
Review Comment:
@dajac @junrao Below is my understand of why the KafkaStorageException was
triggered before this change
1. We explicitly delete the underlying directory via the throwIOException
flag
https://github.com/apache/kafka/blob/184f824cd12ae3e8907ac9066400f2c79fe01d2f/core/src/test/scala/unit/kafka/server/ReplicaManagerTest.scala#L2688
2. During async deletion of the replica, the re-initialization of the leader
epoch cache triggers a FileSystemException
```
<init>:58, IOException (java.io)
<init>:73, FileSystemException (java.nio.file)
translateToIOException:100, UnixException (sun.nio.fs)
rethrowAsIOException:111, UnixException (sun.nio.fs)
rethrowAsIOException:116, UnixException (sun.nio.fs)
newByteChannel:219, UnixFileSystemProvider (sun.nio.fs)
newByteChannel:371, Files (java.nio.file)
createFile:648, Files (java.nio.file)
<init>:104, CheckpointFile (kafka.server.checkpoints)
<init>:68, LeaderEpochCheckpointFile (kafka.server.checkpoints)
newLeaderEpochFileCache$1:2516, Log$ (kafka.log)
maybeCreateLeaderEpochCache:2532, Log$ (kafka.log)
initializeLeaderEpochCache:635, Log (kafka.log)
$anonfun$renameDir$2:758, Log (kafka.log)
renameDir:2699, Log (kafka.log)
asyncDelete:1044, LogManager (kafka.log)
$anonfun$asyncDelete$3:1079, LogManager (kafka.log)
apply:-1, 435995262 (kafka.log.LogManager$$Lambda$842)
foreach:407, Option (scala)
$anonfun$asyncDelete$2$adapted:1077, LogManager (kafka.log)
apply:-1, 1385607480 (kafka.log.LogManager$$Lambda$841)
foreach:79, HashSet (scala.collection.mutable)
asyncDelete:1075, LogManager (kafka.log)
stopPartitions:507, ReplicaManager (kafka.server)
stopReplicas:423, ReplicaManager (kafka.server)
testStopReplicaWithExistingPartition:2378, ReplicaManagerTest (kafka.server)
testStopReplicaWithDeletePartitionAndExistingPartitionAndNewerLeaderEpochAndIOException:2290,
ReplicaManagerTest (kafka.server)
```
3. The FileSystemException is converted into a KafkaStorageException
https://github.com/apache/kafka/blob/eefdf9d6a7fd79a21bb9aea2df25ea642062f28c/core/src/main/scala/kafka/log/LocalLog.scala#L790
Thus when we disable the re-initialization of the leader epoch cache during
the async deletion, the FileSystemException and the KafkaStorageException will
no longer be triggered, and that is what I mean by the underlying directory no
longer being "needed". I can clarify the comment by indicating that the async
deletion code path no longer reads from the underlying directory. Does that
sound ok to you?
--
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: jira-unsubscr...@kafka.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
