Mailing-List: contact hdfs-issues-help@hadoop.apache.org; run by ezmlm
Precedence: bulk
Reply-To: hdfs-issues@hadoop.apache.org
Date: Tue, 14 Jul 2015 02:39:05 +0000 (UTC)
From: "Walter Su (JIRA)" <jira@apache.org>
To: hdfs-issues@hadoop.apache.org
Message-ID: <JIRA.12841862.1435712584000.177997.1436841545079@Atlassian.JIRA>
In-Reply-To: <JIRA.12841862.1435712584000@Atlassian.JIRA>
References: <JIRA.12841862.1435712584000@Atlassian.JIRA>
 <JIRA.12841862.1435712584904@arcas>
Subject: [jira] [Commented] (HDFS-8702) Erasure coding: update
 BlockManager.blockHasEnoughRacks(..) logic for striped block
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit


    [ https://issues.apache.org/jira/browse/HDFS-8702?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14625718#comment-14625718 ] 

Walter Su commented on HDFS-8702:
---------------------------------

bq. We should continue use the passed-in parameter "expectedStorageNum", and we only need to make sure its value is derived from BlockManager#getExpectedReplicaNum which correctly handles striped blocks with different size.

Assume contiguous block:
the pass-in expectedStorageNum only used to test is it need to check enough racks.
{code}
if (expectedStorageNum == 1 || (expectedStorageNum > 1
{code}
See, expectedStorageNum=3 has no difference with expectedStorageNum=4. The expected number of rack, which is 2,  has no relationship with the expectedStorageNum.

As for striped block, expectedStorageNum is useless. Because it always bigger than 1.

1. I think we can use getRealDataBlockNum();
2. activeRacks is useless.
3. It would be better to move the comment to the javadoc of the function.
4. line margin is 80 characters.
5. Indent 4 blank character is enough( Since the function name is long)
{code}
+                                         Collection<DatanodeDescriptor> corruptNodes) {
{code}

> Erasure coding: update BlockManager.blockHasEnoughRacks(..) logic for striped block
> -----------------------------------------------------------------------------------
>
>                 Key: HDFS-8702
>                 URL: https://issues.apache.org/jira/browse/HDFS-8702
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>            Reporter: Walter Su
>            Assignee: Kai Sasaki
>         Attachments: HDFS-8702-HDFS-7285.00.patch, HDFS-8702-HDFS-7285.01.patch, HDFS-8702-HDFS-7285.02.patch
>
>
> Currently blockHasEnoughRacks(..) only guarantees 2 racks. Logic needs updated for striped blocks.


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