[ https://issues.apache.org/jira/browse/HBASE-14004?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15049947#comment-15049947 ]
Heng Chen commented on HBASE-14004: ----------------------------------- {quote} 1, 2, 3, 4, 5---this is normal order 1, 3, 2, 4, 5---the order is wrong but each log we only read once. 1, 1, 2, 3, 4, 5---we replay one entry twice but they are continuous 1, 2, 3, 1, 4, 5----we replay one entry twice and they are not continuous 1, 2, 3, 1, 2, 3, 4, 5---the order is wrong but the subsequence is repeat so we make sure the order {quote} [~yangzhe1991] only the first one is right now in current logic. WAL is RS Level, but replay is in Region Level, so in wal , seqId in increased one by one, but we can't ensure it in region Level when replay, that's why i mentioned i made a mistake in HBASE-14949. I will dig it deeper. Would you mind update the doc as we mentioned above, Zhe? > [Replication] Inconsistency between Memstore and WAL may result in data in > remote cluster that is not in the origin > ------------------------------------------------------------------------------------------------------------------- > > Key: HBASE-14004 > URL: https://issues.apache.org/jira/browse/HBASE-14004 > Project: HBase > Issue Type: Bug > Components: regionserver > Reporter: He Liangliang > Priority: Critical > Labels: replication, wal > > 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 > 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. As a > result, the handler will rollback the Memstore and the later flushed HFile > will also skip this record. -- This message was sent by Atlassian JIRA (v6.3.4#6332)