Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
MapReduce, mail # user - Re: Throttle replication speed in case of datanode failure


Copy link to this message
-
Re: Throttle replication speed in case of datanode failure
Brennon Church 2013-01-17, 21:44
Pretty spiky.  I'll throttle it back to 1MB/s and see if it reduces
things as expected.

Thanks!

--Brennon

On 1/17/13 1:41 PM, Harsh J wrote:
> Not true per the sources, it controls all DN->DN copy/move rates,
> although the property name is misleading. Are you noticing a
> consistent rise in the rate or is it spiky?
>
>
> On Fri, Jan 18, 2013 at 2:20 AM, Brennon Church <[EMAIL PROTECTED]
> <mailto:[EMAIL PROTECTED]>> wrote:
>
>     That doesn't seem to work for under-replicated blocks such as when
>     decommissioning (or losing) a node, just for the balancer.  I've
>     got mine currently set to 10MB/s, but am seeing rates of 3-4 times
>     that after decommissioning a node while it works on bringing
>     things back up to the proper replication factor.
>
>     Thanks.
>
>     --Brennon
>
>
>     On 1/17/13 11:04 AM, Harsh J wrote:
>>     You can limit the bandwidth in bytes/second values applied
>>     via dfs.balance.bandwidthPerSec in each DN's hdfs-site.xml.
>>     Default is 1 MB/s (1048576).
>>
>>     Also, unsure if your version already has it, but it can be
>>     applied at runtime too via the dfsadmin -setBalancerBandwidth
>>     command.
>>
>>
>>     On Thu, Jan 17, 2013 at 8:11 PM, Brennon Church
>>     <[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>> wrote:
>>
>>         Hello,
>>
>>         Is there a way to throttle the speed at which
>>         under-replicated blocks are copied across a cluster?  Either
>>         limiting the bandwidth or the number of blocks per time
>>         period would work.
>>
>>         I'm currently running Hadoop v1.0.1.  I think the
>>         dfs.namenode.replication.work.multiplier.per.iteration option
>>         would do the trick, but that is in v1.1.0 and higher.
>>
>>         Thanks.
>>
>>         --Brennon
>>
>>
>>
>>
>>     --
>>     Harsh J
>
>
>
>
> --
> Harsh J