[
https://issues.apache.org/jira/browse/HADOOP-4477?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12642279#action_12642279
]
Hairong Kuang commented on HADOOP-4477:
---------------------------------------
My proposal is to include both under-replicated blocks and blocks that do not
satisfy rack requirement in the neededReplication queue. The neededReplication
queue supports four priorities:
Priority 0: Blocks that have only one replicas;
Priority 1: Blocks whose replicas are on only one rack;
Priority 2: Blocks whose number of replicas is no greater than 1/3 of it
replication factor;
Priority 3: All other under-replicated blocks.
In general we should have priority 4 which includes those blocks that do not
belong to priorities 0-3 and do not satisfy the HDFS rack requirement.
Currently HDFS provides only two-rack guarantee so priority 1 covers all rack
requirement break cases.
In methods addStoredBlock, removeStoredBlock, startDecomission, and
markBlockAsCorrupt in FSNamesystem, put both under-replication and 1 rack
blocks into the neededReplication queue. Replicator will in addition replicate
one more replicas for only 1 rack not under-replicated blocks.
> All replicas of a block end up on only 1 rack
> ---------------------------------------------
>
> Key: HADOOP-4477
> URL: https://issues.apache.org/jira/browse/HADOOP-4477
> Project: Hadoop Core
> Issue Type: Bug
> Components: dfs
> Reporter: Hairong Kuang
> Assignee: Hairong Kuang
> Priority: Critical
> Fix For: 0.20.0
>
>
> HDFS replicas placement strategy guarantees that the replicas of a block
> exist on at least two racks when its replication factor is greater than one.
> But fsck still reports that the replicas of some blocks end up on one rack.
> The cause of the problem is that decommission and corruption handling only
> check the block's replication factor but not the rack requirement. When an
> over-replicated block loses a replica due to decomission, corruption, or
> heartbeat lost, namenode does not take any action to guarantee that remaining
> replicas are on different racks.
>
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
