hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ming Ma (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-9314) Improve BlockPlacementPolicyDefault's picking of excess replicas
Date Thu, 19 Nov 2015 23:25:11 GMT

    [ https://issues.apache.org/jira/browse/HDFS-9314?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15014734#comment-15014734
] 

Ming Ma commented on HDFS-9314:
-------------------------------

Thanks [~xiaochen]. This should work.

* The rack-to-hosts mapping {{rackMap}} has already been computed. Do you think it is clearer
if we modify {{pickupReplicaSet}} to take the rack-to-hosts mapping {{rackMap}} as input instead?
In that way, it can say something like: a) if number of existing replicas is 2, both nodes
are candidates. b) if number of racks >=3, all nodes are candidates. c) if the number of
racks <=2, all nodes that share racks with at least another node are candidates.

* As far as I can tell, this shouldn't impact {{BlockPlacementPolicyWithUpgradeDomain}}. It
will be useful if you can add a new test case to {{TestReplicationPolicyWithUpgradeDomain}}
to make sure it works. Also the description for {{BlockPlacementPolicyWithUpgradeDomain}}'s
{{pickupReplicaSet}} will need some update as {{BlockPlacementPolicyDefault}}'s {{pickupReplicaSet}}
no longer only return set of share-rack nodes anymore; instead it will return a super set
of share-rack nodes.

> Improve BlockPlacementPolicyDefault's picking of excess replicas
> ----------------------------------------------------------------
>
>                 Key: HDFS-9314
>                 URL: https://issues.apache.org/jira/browse/HDFS-9314
>             Project: Hadoop HDFS
>          Issue Type: Improvement
>            Reporter: Ming Ma
>            Assignee: Xiao Chen
>         Attachments: HDFS-9314.001.patch, HDFS-9314.002.patch, HDFS-9314.003.patch
>
>
> The test case used in HDFS-9313 identified NullPointerException as well as the limitation
of excess replica picking. If the current replicas are on {SSD(rack r1), DISK(rack 2), DISK(rack
3), DISK(rack 3)} and the storage policy changes to HOT_STORAGE_POLICY_ID, BlockPlacementPolicyDefault's
won't be able to delete SSD replica.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message