[
https://issues.apache.org/jira/browse/HDFS-15725?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17247501#comment-17247501
]
Stephen O'Donnell commented on HDFS-15725:
------------------------------------------
Yes, it is probably the shutdown hook that is ultimately causing this, along
with perhaps Kafka Connect not shutting down cleanly (I have not tried to
confirm that).
NN logs:
{code}
2020-12-09 08:00:01,773 DEBUG org.apache.hadoop.hdfs.StateChange:
persistBlocks: /redacted/file with 1 blocks is peristed to the file system
2020-12-09 08:00:01,831 DEBUG org.apache.hadoop.hdfs.StateChange: *DIR*
NameNode.complete: /redacted/file fileId=3288929 for
DFSClient_NONMAPREDUCE_374148706_65
2020-12-09 08:00:01,831 DEBUG org.apache.hadoop.hdfs.StateChange: DIR*
NameSystem.completeFile: /redacted/file for DFSClient_NONMAPREDUCE_374148706_65
2020-12-09 08:00:01,831 INFO
org.apache.hadoop.hdfs.server.namenode.FSNamesystem: BLOCK* checkFileProgress:
blk_1075927944_2211851{blockUCState=COMMITTED, primaryNodeIndex=-1,
replicas=[ReplicaUnderConstruction[[DISK]DS-ab35a9f9-9439-4b05-9ea1-9d68e0c5f11f:NORMAL:10.32.136.71:1044|RBW],
ReplicaUnderConstruction[[DISK]DS-d07f48a7-258c-484a-af96-010557f70fa8:NORMAL:10.32.136.62:1044|RBW]]}
has not reached minimal replication 1
<nothing further until next client tries to write it>
{code}
DN logs (all 3 DNs are basically the same):
{code}
{"type":"log","host":"host_name","category":"HDFS-hdfs-DATANODE-BASE","level":"DEBUG","system":"HEKODLK","time":
"2020-12-09
08:00:01,822","logger":"datanode.DataNode","timezone":"EST","log":{"message":"PacketResponder:
BP-1714193774-10.32.136.20-1601039015398:blk_1075927944_2211851,
type=LAST_IN_PIPELINE, downstreams=0:[]: seqno=-2 waiting for local datanode to
finish write."}}
{"type":"log","host":"host_name","category":"HDFS-hdfs-DATANODE-BASE","level":"INFO","system":"HEKODLK","time":
"2020-12-09
08:00:01,822","logger":"datanode.DataNode","timezone":"EST","log":{"message":"Exception
for BP-1714193774-10.32.136.20-1601039015398:blk_1075927944_2211851"}}
{"type":"log","host":"host_name","category":"HDFS-hdfs-DATANODE-BASE","level":"INFO","system":"HEKODLK","time":
"2020-12-09
08:00:01,823","logger":"datanode.DataNode","timezone":"EST","log":{"message":"PacketResponder:
BP-1714193774-10.32.136.20-1601039015398:blk_1075927944_2211851,
type=LAST_IN_PIPELINE, downstreams=0:[]: Thread is interrupted."}}
{"type":"log","host":"host_name","category":"HDFS-hdfs-DATANODE-BASE","level":"INFO","system":"HEKODLK","time":
"2020-12-09
08:00:01,823","logger":"datanode.DataNode","timezone":"EST","log":{"message":"PacketResponder:
BP-1714193774-10.32.136.20-1601039015398:blk_1075927944_2211851,
type=LAST_IN_PIPELINE, downstreams=0:[] terminating"}}
{"type":"log","host":"host_name","category":"HDFS-hdfs-DATANODE-BASE","level":"INFO","system":"HEKODLK","time":
"2020-12-09
08:00:01,823","logger":"datanode.DataNode","timezone":"EST","log":{"message":"opWriteBlock
BP-1714193774-10.32.136.20-1601039015398:blk_1075927944_2211851 received
exception java.io.IOException: Connection reset by peer"}}
{code}
Notice the DN gets the Connection Reset shortly after the NN, so this probably
is the shutdown hook on the client. It was killed over and over until this
happened.
> Lease Recovery never completes for a committed block which the DNs never
> finalize
> ---------------------------------------------------------------------------------
>
> Key: HDFS-15725
> URL: https://issues.apache.org/jira/browse/HDFS-15725
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: namenode
> Affects Versions: 3.4.0
> Reporter: Stephen O'Donnell
> Assignee: Stephen O'Donnell
> Priority: Major
> Attachments: HDFS-15725.001.patch, lease_recovery_2_10.patch
>
>
> It a very rare condition, the HDFS client process can get killed right at the
> time it is completing a block / file.
> The client sends the "complete" call to the namenode, moving the block into a
> committed state, but it dies before it can send the final packet to the
> Datanodes telling them to finalize the block.
> This means the blocks are stuck on the datanodes in RBW state and nothing
> will ever tell them to move out of that state.
> The namenode / lease manager will retry forever to close the file, but it
> will always complain it is waiting for blocks to reach minimal replication.
> I have a simple test and patch to fix this, but I think it warrants some
> discussion on whether this is the correct thing to do, or if I need to put
> the fix behind a config switch.
> My idea, is that if lease recovery occurs, and the block is still waiting on
> "minimal replication", just put the file back to UNDER_CONSTRUCTION so that
> on the next lease recovery attempt, BLOCK RECOVERY will happen, close the
> file and move the replicas to FINALIZED.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
