Details
-
Type:
Bug
-
Status: Resolved
-
Priority:
Major
-
Resolution: Fixed
-
Affects Version/s: None
-
Fix Version/s: 1.6.0
-
Component/s: None
-
Labels:None
Description
"The issue is that the file close fails, but the stream on the Flume side is considered closed by HDFS. So retrying the close, is failing - which causes the rename never to be attempted. Since the rename is never attempted the file sticks around with the ".tmp" file name. "
We should try attempting rename even if the close fails and possibly log information that the user need to run lease recovery
Issue Links
- relates to
-
FLUME-2566
BucketWriter tries to close file endlessly
-
- Closed
-
Activity
- All
- Comments
- Work Log
- History
- Activity
- Transitions
Johny Rufus, The patch looks good. I am running tests. If all tests pass, I will commit this.
+1. Committing.
Commit 0d6eccad2781884fe0f000f74b8f964cbdd7971f in flume's branch refs/heads/trunk from Hari Shreedharan
[ https://git-wip-us.apache.org/repos/asf?p=flume.git;h=0d6ecca ]
FLUME-2586. HDFS Sink must try to rename files even if close fails.
(Johny Rufus via Hari)
Commit 13a03003e7c1817d73451d3b6b1a58250c9469bd in flume's branch refs/heads/flume-1.6 from Hari Shreedharan
[ https://git-wip-us.apache.org/repos/asf?p=flume.git;h=13a0300 ]
FLUME-2586. HDFS Sink must try to rename files even if close fails.
(Johny Rufus via Hari)
Committed! Thanks Johny!
UNSTABLE: Integrated in flume-trunk #709 (See https://builds.apache.org/job/flume-trunk/709/)
FLUME-2586. HDFS Sink must try to rename files even if close fails. (hshreedharan: http://git-wip-us.apache.org/repos/asf/flume/repo?p=flume.git&a=commit&h=0d6eccad2781884fe0f000f74b8f964cbdd7971f)
- flume-ng-sinks/flume-hdfs-sink/src/main/java/org/apache/flume/sink/hdfs/BucketWriter.java
- flume-ng-sinks/flume-hdfs-sink/src/test/java/org/apache/flume/sink/hdfs/MockFsDataOutputStream.java
- flume-ng-sinks/flume-hdfs-sink/src/test/java/org/apache/flume/sink/hdfs/TestBucketWriter.java
- flume-ng-sinks/flume-hdfs-sink/src/test/java/org/apache/flume/sink/hdfs/TestHDFSEventSink.java
- flume-ng-sinks/flume-hdfs-sink/src/test/java/org/apache/flume/sink/hdfs/MockFileSystem.java
UNSTABLE: Integrated in Flume-trunk-hbase-98 #66 (See https://builds.apache.org/job/Flume-trunk-hbase-98/66/)
FLUME-2586. HDFS Sink must try to rename files even if close fails. (hshreedharan: http://git-wip-us.apache.org/repos/asf/flume/repo?p=flume.git&a=commit&h=0d6eccad2781884fe0f000f74b8f964cbdd7971f)
- flume-ng-sinks/flume-hdfs-sink/src/test/java/org/apache/flume/sink/hdfs/MockFsDataOutputStream.java
- flume-ng-sinks/flume-hdfs-sink/src/test/java/org/apache/flume/sink/hdfs/MockFileSystem.java
- flume-ng-sinks/flume-hdfs-sink/src/main/java/org/apache/flume/sink/hdfs/BucketWriter.java
- flume-ng-sinks/flume-hdfs-sink/src/test/java/org/apache/flume/sink/hdfs/TestBucketWriter.java
- flume-ng-sinks/flume-hdfs-sink/src/test/java/org/apache/flume/sink/hdfs/TestHDFSEventSink.java
Johny Rufus Hari Shreedharan looks like this feature went undocumented.
It just reuses an older config param, "hdfs.closeTries" - we just need to update the description I guess.
Commit 4188e4dedb4af0792bfac3086e3b2a7d415f2248 in flume's branch refs/heads/trunk from Hari Shreedharan
[ https://git-wip-us.apache.org/repos/asf?p=flume.git;h=4188e4d ]
FLUME-2665. Update documentation for hdfs.closeTries based on FLUME-2586
(Johny Rufus via Hari)
Commit dbf4828b89a36fda9434163f1835382e2f0c56ba in flume's branch refs/heads/flume-1.6 from Hari Shreedharan
[ https://git-wip-us.apache.org/repos/asf?p=flume.git;h=dbf4828 ]
FLUME-2665. Update documentation for hdfs.closeTries based on FLUME-2586
(Johny Rufus via Hari)
SUCCESS: Integrated in Flume-trunk-hbase-98 #90 (See https://builds.apache.org/job/Flume-trunk-hbase-98/90/)
FLUME-2665. Update documentation for hdfs.closeTries based on FLUME-2586 (hshreedharan: http://git-wip-us.apache.org/repos/asf/flume/repo?p=flume.git&a=commit&h=4188e4dedb4af0792bfac3086e3b2a7d415f2248)
- flume-ng-doc/sphinx/FlumeUserGuide.rst
Calling close multiple times on the HDFSStream has no effect, hence in this patch, the closeTries parameter is used to retry rename, (irrespective of wether the close succeeded or not)