[
https://issues.apache.org/jira/browse/HDFS-16076?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17367252#comment-17367252
]
tomscut commented on HDFS-16076:
--------------------------------
The main changes are as follows:
1 Change slowNodesSet into slowNodesUuidSet to improve the universality of slow
node set. Because datanodeUuid is more useful for identifying unique DataNode.
2 In FSNamesystem#getBlockLocations(), place the slow node as far back as
possible to reduce the probability that the slow node will be used for reading.
3 The ability to avoid using slow nodes is configurable.
> Avoid using slow DataNodes for reading by sorting locations
> -----------------------------------------------------------
>
> Key: HDFS-16076
> URL: https://issues.apache.org/jira/browse/HDFS-16076
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: hdfs
> Reporter: tomscut
> Assignee: tomscut
> Priority: Major
> Labels: pull-request-available
> Time Spent: 1h 10m
> Remaining Estimate: 0h
>
> After sorting the expected location list will be: live -> slow -> stale ->
> staleAndSlow -> entering_maintenance -> decommissioned. This reduces the
> probability that slow nodes will be used for reading.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
