He Liangliang created HBASE-14004:
-------------------------------------
Summary: Inconsistency between Memstore and WAL
Key: HBASE-14004
URL: https://issues.apache.org/jira/browse/HBASE-14004
Project: HBase
Issue Type: Bug
Components: regionserver
Reporter: He Liangliang
Priority: Critical
We encountered data inconsistency issue between master and slave clusters.
Looks like the current write path can cause inconsistency between
memstore/hfile and WAL which cause the slave cluster has more data than the
master cluster.
The simplified write path looks like:
1. insert record into Memstore
2. write record to WAL
3. sync WAL
4. rollback Memstore if 3 fails
In step 3, the sync operation can fail in two case:
1. The caller disconnected the socket due to timeout or any other client crashes
2. The underlying HDFS sync operation can fails
For case 1, the caller may disconnect the socket (for example, timeout due to
slow sync), but the sync can finally succeed because it's an asynchronous
operation. As a result, the handler will rollback the Memstore and the later
flushed HFile will also skip this record.
For case 2, it's possible that the HDFS sync RPC call fails, but the data is
already (may partially) transported to the DNs which finally get persisted.
0.94 code fails in both cases, and looks like the trunk code suffered only case
2.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
