hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Zhe Zhang (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-9381) When same block came for replication for Striped mode, we can move that block to PendingReplications
Date Tue, 01 Dec 2015 00:34:11 GMT

    [ https://issues.apache.org/jira/browse/HDFS-9381?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15032807#comment-15032807

Zhe Zhang commented on HDFS-9381:

Thanks Jing for the comment.

Let's consider this case:
# Cluster has 100 nodes
# DN_1 and DN_2 failed
# They are on different racks
# They happen to share 1000 striped blocks and 1000 contiguous blocks (we can easily scale
up the calculated numbers for n x 1000 blocks). So there are 2000 striped internal blocks,
and 2000 contiguous block replicas missing.

So in each iteration ReplicationMonitor tries to pickup 200 items. Without the change, it
will be 100 striped and 100 contiguous on average. Assuming EC recovery work takes longer
than 3 seconds ({{replicationRecheckInterval}}), then the 2nd iteration will pickup about
5 invalid striped items (being recovered). If EC recovery work takes long enough, then the
3rd round will pickup about 10 (2/18) invalid striped items, 4th round 18 invalid items. This
way the replication work for the lost contiguous replicas will take 20 x 3 = 30 seconds to
distributed to DNs.

With the change, 2nd round will pickup 95 striped items and 105 contiguous items, 3rd round
110 contiguous items, .... It's tricky to get very accurate, but seems we can save a few 3-second

[~umamaheswararao] Does the example itself make sense to you? If so, how should we calculate
the saving in locking time?

bq. it is also possible that because of the longer processing time, there is higher chance
for the striped blocks to be updated in the UC queue before being processed by the replication
monitor for the first time
I'm not fully following the above. Jing do you mind elaborating it a little bit?

> When same block came for replication for Striped mode, we can move that block to PendingReplications
> ----------------------------------------------------------------------------------------------------
>                 Key: HDFS-9381
>                 URL: https://issues.apache.org/jira/browse/HDFS-9381
>             Project: Hadoop HDFS
>          Issue Type: Sub-task
>          Components: erasure-coding, namenode
>    Affects Versions: 3.0.0
>            Reporter: Uma Maheswara Rao G
>            Assignee: Uma Maheswara Rao G
>         Attachments: HDFS-9381-02.patch, HDFS-9381-03.patch, HDFS-9381.00.patch, HDFS-9381.01.patch
> Currently I noticed that we are just returning null if block already exists in pendingReplications
in replication flow for striped blocks.
> {code}
> if (block.isStriped()) {
>       if (pendingNum > 0) {
>         // Wait the previous recovery to finish.
>         return null;
>       }
> {code}
>  Here if we just return null and if neededReplications contains only fewer blocks(basically
by default if less than numliveNodes*2), then same blocks can be picked again from neededReplications
from next loop as we are not removing element from neededReplications. Since this replication
process need to take fsnamesystmem lock and do, we may spend some time unnecessarily in every
> So my suggestion/improvement is:
>  Instead of just returning null, how about incrementing pendingReplications for this
block and remove from neededReplications? and also another point to consider here is, to add
into pendingReplications, generally we need target and it is nothing but to which node we
issued replication command. Later when after replication success and DN reported it, block
will be removed from pendingReplications from NN addBlock. 
>  So since this is newly picked block from neededReplications, we would not have selected
target yet. So which target to be passed to pendingReplications if we add this block? One
Option I am thinking is, how about just passing srcNode itself as target for this special
condition? So, anyway if the block is really missed, srcNode will not report it. So this block
will not be removed from pending replications, so that when it is timed out, it will be considered
for replication again and that time it will find actual target to replicate while processing
as part of regular replication flow.

This message was sent by Atlassian JIRA

View raw message