[ https://issues.apache.org/jira/browse/HBASE-5074?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13215108#comment-13215108 ]
Phabricator commented on HBASE-5074: ------------------------------------ dhruba has commented on the revision "[jira] [HBASE-5074] Support checksums in HBase block cache". INLINE COMMENTS src/main/java/org/apache/hadoop/hbase/io/hfile/HFileBlock.java:235 This constructor is used only for V2, hence the major number is not a parameter. src/main/java/org/apache/hadoop/hbase/io/hfile/HFileBlock.java:1235 I think there won;t be any changes to the number of threads in the datanode. A datanode thread is not tied up with a client FileSystem object. Instead, a global pool of threads in the datanode are free to serve any read-requests from any client src/main/java/org/apache/hadoop/hbase/io/hfile/HFileBlock.java:1244 The minor version indicates disk-format changes inside an HFileBlock. The major version indicates disk-format changes within a entire HFile. Since the AbstractFSReader only reads HFileBlocks, so it is logical that it contains the minorVersion, is it not? But I can put in the majorVersion in it as well, if you so desire. src/main/java/org/apache/hadoop/hbase/io/hfile/HFileBlock.java:1584 Yes, the default it to enable hbase-checksum verification. And you are right that if the hfile is of the older type, then we will quickly flip this back to false (in the next line) src/main/java/org/apache/hadoop/hbase/io/hfile/HFileBlock.java:1630 I think we should keep both streams active till the HFile itself is closed. src/main/java/org/apache/hadoop/hbase/io/hfile/HFileBlock.java:1646 done src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java:961 Yes, precisely. Going forward, I would like to see if we can make HLogs go to a filesystem object that is different from the filesystem used for hfiles. src/main/java/org/apache/hadoop/hbase/regionserver/RegionServerServices.java:86 I agree with you completely. This is an interface that should not change often. REVISION DETAIL https://reviews.facebook.net/D1521 > support checksums in HBase block cache > -------------------------------------- > > Key: HBASE-5074 > URL: https://issues.apache.org/jira/browse/HBASE-5074 > Project: HBase > Issue Type: Improvement > Components: regionserver > Reporter: dhruba borthakur > Assignee: dhruba borthakur > Attachments: D1521.1.patch, D1521.1.patch, D1521.2.patch, > D1521.2.patch, D1521.3.patch, D1521.3.patch, D1521.4.patch, D1521.4.patch, > D1521.5.patch, D1521.5.patch, D1521.6.patch, D1521.6.patch, D1521.7.patch, > D1521.7.patch, D1521.8.patch, D1521.8.patch > > > The current implementation of HDFS stores the data in one block file and the > metadata(checksum) in another block file. This means that every read into the > HBase block cache actually consumes two disk iops, one to the datafile and > one to the checksum file. This is a major problem for scaling HBase, because > HBase is usually bottlenecked on the number of random disk iops that the > storage-hardware offers. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira