[ https://issues.apache.org/jira/browse/HDFS-2123?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13060046#comment-13060046 ]
Bharath Mundlapudi commented on HDFS-2123: ------------------------------------------ I have reviewed the patch. This change is more meaningful than check pointing based on size. +1 to the approach. There are some minor comments on logging msgs since we are moving to txns, we should reflect this in log msgs too. 1. Replace in Checkpointer + LOG.info("Log Size Trigger : " + checkpointTxnCount + " txns "); With + LOG.info("Transaction Count Trigger : " + checkpointTxnCount + " txns "); 2. Replace in SecondaryNameNode + + "\nCheckpoint Size : " + StringUtils.byteDesc(checkpointTxnCount) + + " (= " + checkpointTxnCount + " bytes)" With + + "\nTransaction Count : " + StringUtils.byteDesc(checkpointTxnCount) + + " (= " + checkpointTxnCount + " txns)" 3. Replace in SecondaryNamenode + LOG.info("Log Size Trigger :" + checkpointTxnCount + " txns"); with + LOG.info("Transaction Count Trigger : " + checkpointTxnCount + " txns "); 4. Replace in SecondaryNamenode + System.err.println("EditLog size " + count + " transactions is " + "smaller than configured checkpoint " + + "interval " + checkpointTxnCount + " transactions."); with + System.err.println("EditLog transactions " + count + " is " + "smaller than configured checkpoint " + + "transactions " + checkpointTxnCount); > 1073: Checkpoint interval should be based on txn count, not size > ---------------------------------------------------------------- > > Key: HDFS-2123 > URL: https://issues.apache.org/jira/browse/HDFS-2123 > Project: Hadoop HDFS > Issue Type: Sub-task > Components: name-node > Affects Versions: Edit log branch (HDFS-1073) > Reporter: Todd Lipcon > Assignee: Todd Lipcon > Fix For: Edit log branch (HDFS-1073) > > Attachments: hdfs-2123.txt, hdfs-2123.txt > > > Currently, the administrator can configure the secondary namenode to > checkpoint either every N seconds, or every N bytes worth of edit log. It > would make more sense to get rid of the size-based interval and instead allow > the administrator to specify checkpoints every N transactions. This also > simplifies the code a little bit. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira