[
https://issues.apache.org/jira/browse/HDFS-8872?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15317548#comment-15317548
]
Ming Ma commented on HDFS-8872:
-------------------------------
Thanks [~shahrs87]. Your suggestion to distinguish decommissioned from
decommissioning makes sense. Given liveness state is separated from admin
state, it appears the suggestion is "a block should be considered missing if
and only if {live, normal} + {live, decommissioning} == 0". The live attribute
is implicit given fsck only gets the list of live replicas; but it might be
useful to call it out.
If that is the case, both fsck and jmx metrics need to be changed. Also
depending on how you interpret it, it seems more like incompatible change than
just bug fix. That will decide if we want to fix it only in trunk.
> Reporting of missing blocks is different in fsck and namenode ui/metasave
> -------------------------------------------------------------------------
>
> Key: HDFS-8872
> URL: https://issues.apache.org/jira/browse/HDFS-8872
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 2.6.0
> Reporter: Rushabh S Shah
> Assignee: Rushabh S Shah
>
> Namenode ui and metasave will not report a block as missing if the only
> replica is on decommissioning/decomissioned node while fsck will show it as
> MISSING.
> Since decommissioned node can be formatted/removed anytime, we can actually
> lose the block.
> Its better to alert on namenode ui if the only copy is on
> decomissioned/decommissioning node.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
