ZanderXu created HDFS-16659:
-------------------------------
Summary: JournalNode should throw CacheMissException when
SinceTxId is more than HighestWrittenTxId
Key: HDFS-16659
URL: https://issues.apache.org/jira/browse/HDFS-16659
Project: Hadoop HDFS
Issue Type: Bug
Reporter: ZanderXu
Assignee: ZanderXu
JournalNode should throw `CacheMissException` if `sinceTxId` is bigger than
`highestWrittenTxId`. And it will caused EditlogTailer can not able to tail
edits. And it maybe caused ObserverNameNode can not able handle requests from
clients.
Suppose there are 3 journalNodes, JN0 ~ JN1.
The corner case as blew:
* JN0 has some abnormal cases when Active Namenode is journaling Edits with
start txId 11
* NameNode just ignore the abnormal JN0 and continue to write Edits to Journal
1 and 2
* JN0 backed to health
* Observer NameNode try to select EditLogInputStream vis PRC with start txId 21
* Journal 1 has some abnormal cases caused slow rpc response
And the expected selecting result is: Response should contain 20 Edits from
txId 21 to txId 40 from JN1 and JN2. Because Active NameNode successfully write
these Edits to JN1 and JN2 and failed write these edits to JN0, so there is no
Edits from id 21 to 40 in the cache of JN0.
But in the current implementation, there is no Edits in the Response. Because
namenode successfully got a response from JN0 that did not contains any Edits.
And the bug code as blew:
{code:java}
if (sinceTxId > getHighestWrittenTxId()) {
// Requested edits that don't exist yet; short-circuit the cache here
metrics.rpcEmptyResponses.incr();
return GetJournaledEditsResponseProto.newBuilder().setTxnCount(0).build();
}
{code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-dev-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-dev-h...@hadoop.apache.org
