hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "nkeywal (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-3703) Decrease the datanode failure detection time
Date Mon, 23 Jul 2012 18:30:35 GMT

    [ https://issues.apache.org/jira/browse/HDFS-3703?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13420851#comment-13420851
] 

nkeywal commented on HDFS-3703:
-------------------------------

bq. Can you describe this better?
If we see this in layers, we've got three layers:
1) Hardware
2) HDFS
3) HBase

Here, the layer3 knows/guess that the layer1 is dead, while the layer in the middle does not
know it. That's not a perfect example of encapsulation :-). HBase is saying to hdfs 'you know,
I want some blocks, but may be this datanode is not good, I'm not sure, but don't use it please'.
Kind of strange (but useful short term).

Today, when there is a global issue, HBase starts its recovery while hdfs is still ignoring
the issue. It leads to a nightmare of socket exception all over the place, as HBase is directed
to dead nodes again and again. HDFS should know before HBase what's going on. So if HBase
it set with a timeout of 30s, HDFS should have 20s or something like this.

bq. Whether ZooKeeper or Datanode heartbeat to Namenode, at a high level mechanisms are similar.


Fully agreed. Just that if the issues comes from ZK or ZK links, HBase and HDFS they would
have a similar view of the situation (may be a wrong view but the same view). On the other
hand, there are possible improvements, not available in ZK, but hopefully available a day,
when there will be more code to share (I'm thinking about ZOOKEEPER-702). Also, still long
term, ZK creates one tcp connection per process monitored. If multiple hadoop processes share
the same tech, it will make sense to have a shared component on each computer to lower the
number of connections. I'm not aware on anything on this subject in ZK, so that's science
fiction today. I've got other stuff like this in mind, but you got the idea :-).

So, I fully agree with your main point, today the real issue is the right timeout.

bq. The problem is one of choosing right timeout. Currently this is configurable in HDFS and
10 minutes is chosen as the timeout. I suggest runningt some experiments with setting this
to a more aggressive value. I agree that this is a very conservative time. But false positives
here could result in replication storm.

Agreed, even we've the current setting, people had issues in the past. 10 minutes seems to
be a reasonable-real world-validated timeout for re-replicating. I don't think it's a good
idea to make lower. However, I think it would be good to have a middle state between fully
available and definitively dead: the non responding nodes could be removed from the target
list for new blocks and de-prioritize for reads.

                
> Decrease the datanode failure detection time
> --------------------------------------------
>
>                 Key: HDFS-3703
>                 URL: https://issues.apache.org/jira/browse/HDFS-3703
>             Project: Hadoop HDFS
>          Issue Type: Improvement
>          Components: data-node, name-node
>    Affects Versions: 1.0.3, 2.0.0-alpha
>            Reporter: nkeywal
>
> By default, if a box dies, the datanode will be marked as dead by the namenode after
10:30 minutes. In the meantime, this datanode will still be proposed  by the nanenode to write
blocks or to read replicas. It happens as well if the datanode crashes: there is no shutdown
hooks to tell the nanemode we're not there anymore.
> It especially an issue with HBase. HBase regionserver timeout for production is often
30s. So with these configs, when a box dies HBase starts to recover after 30s and, while 10
minutes, the namenode will consider the blocks on the same box as available. Beyond the write
errors, this will trigger a lot of missed reads:
> - during the recovery, HBase needs to read the blocks used on the dead box (the ones
in the 'HBase Write-Ahead-Log')
> - after the recovery, reading these data blocks (the 'HBase region') will fail 33% of
the time with the default number of replica, slowering the data access, especially when the
errors are socket timeout (i.e. around 60s most of the time). 
> Globally, it would be ideal if HDFS settings could be under HBase settings. 
> As a side note, HBase relies on ZooKeeper to detect regionservers issues.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message