[
https://issues.apache.org/jira/browse/HDFS-3077?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13473783#comment-13473783
]
Sanjay Radia commented on HDFS-3077:
------------------------------------
bq. If a JN crashes and is reformatted, then this would imply that the JN has
to copy multiple GB worth of data from another JN before it can actively start
participating as a destination for new logs. This will take quite some time.
* Correctness - unless i am missing something, as soon as one of the JN's disk
is reformatted you have lost the property that each segment is replicated at at
least Q JNs. Hence it is best to recover all segments on this reformatted JN.
* Performance: The QJM client does parallel writes so a single slow JN will not
be a problem. Further, NN batches journal write and hence the batching will
increase if QJM slows down.
* Comparison with local disk: Local disks fail much less than connections to
JNs in a busy network. Hence you are likely to see more segments and more gaps
in QJM compared to local disk based journal.
But we can discuss these further and in the worst case we can make the
full-vs-partial recovery configurable.
Please proceed with the merge. Thanks for the work you put into this Jira.
> Quorum-based protocol for reading and writing edit logs
> -------------------------------------------------------
>
> Key: HDFS-3077
> URL: https://issues.apache.org/jira/browse/HDFS-3077
> Project: Hadoop HDFS
> Issue Type: New Feature
> Components: ha, name-node
> Reporter: Todd Lipcon
> Assignee: Sanjay Radia
> Fix For: QuorumJournalManager (HDFS-3077)
>
> Attachments: hdfs-3077-partial.txt, hdfs-3077-test-merge.txt,
> hdfs-3077.txt, hdfs-3077.txt, hdfs-3077.txt, hdfs-3077.txt, hdfs-3077.txt,
> hdfs-3077.txt, hdfs-3077.txt, qjournal-design.pdf, qjournal-design.pdf,
> qjournal-design.pdf, qjournal-design.pdf, qjournal-design.pdf,
> qjournal-design.pdf, qjournal-design.tex, qjournal-design.tex
>
>
> Currently, one of the weak points of the HA design is that it relies on
> shared storage such as an NFS filer for the shared edit log. One alternative
> that has been proposed is to depend on BookKeeper, a ZooKeeper subproject
> which provides a highly available replicated edit log on commodity hardware.
> This JIRA is to implement another alternative, based on a quorum commit
> protocol, integrated more tightly in HDFS and with the requirements driven
> only by HDFS's needs rather than more generic use cases. More details to
> follow.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
