Mailing-List: contact hdfs-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: hdfs-issues@hadoop.apache.org
Date: Tue, 17 Nov 2015 02:17:11 +0000 (UTC)
From: "Walter Su (JIRA)" <jira@apache.org>
To: hdfs-issues@hadoop.apache.org
Message-ID: <JIRA.12908085.1445909088000.88196.1447726631028@Atlassian.JIRA>
In-Reply-To: <JIRA.12908085.1445909088000@Atlassian.JIRA>
References: <JIRA.12908085.1445909088000@Atlassian.JIRA>
 <JIRA.12908085.1445909088376@arcas>
Subject: [jira] [Commented] (HDFS-9314) Improve
 BlockPlacementPolicyDefault's picking of excess replicas
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HDFS-9314?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15007848#comment-15007848 ] 

Walter Su commented on HDFS-9314:
---------------------------------

Shall we consider not to violate the placement policy when we decide to decrease rack counts?
In [HDFS-9313|https://issues.apache.org/jira/browse/HDFS-9313?focusedCommentId=14975592&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-14975592] I agree to choose the SSD, since the remaining 3 DISKs are already on enough racks.

Consider a different case,
||rack_0||rack_1||
|replica_0(DISK)|replica_3(SSD)|
|replica_1(DISK)|
|replica_2(DISK)|
If we remove replica_3, the left are not on enough rack. Is it removing replica_3 the right decision any more? Shall we make sure not to violate the placement policy, before decide to decrease rack counts? We can schedule a replication task first, then come back to delete the SSD.
The old logic will choose from {{exactlyOne}} collection as well, but only if {{moreThanOne}} is empty. So the old logic doesn't violate the placement policy.

> 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
>
>
> 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 1), DISK(rack 2), DISK(rack 2)} 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)