[
https://issues.apache.org/jira/browse/HDFS-11482?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chen Liang updated HDFS-11482:
------------------------------
Attachment: HDFS-11482.005.patch
Thanks [~linyiqun] for the review and the comments! All the comments except the
following one are addressed in v005 patch.
bq. If excludeRoot is a DatanodeDescriptor and it has more than one storage of
give type, is right to just subtract 1?
The counts are not "how many storages of this type there are" but "how many
datanodes have this storage type". The idea is that, all we need is to find a
datanode that has the storage type. Exactly how many it has does not matter. So
if a datanode has that type, it is counted as "1", although it may have many of
that storage. Similarly when excluding a datanode, it subtract by "1" as well.
> Add storage type demand to into DFSNetworkTopology#chooseRandom
> ---------------------------------------------------------------
>
> Key: HDFS-11482
> URL: https://issues.apache.org/jira/browse/HDFS-11482
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: namenode
> Reporter: Chen Liang
> Assignee: Chen Liang
> Attachments: HDFS-11482.001.patch, HDFS-11482.002.patch,
> HDFS-11482.003.patch, HDFS-11482.004.patch, HDFS-11482.005.patch
>
>
> HDFS-11450 adds storage type info into network topology, with on this info we
> may change chooseRandom to take storage type requirement as parameter, only
> checking subtrees required storage type available. This way we avoid blindly
> picking up nodes that are not applicable.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
