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

Switch to Threaded View
HBase, mail # user - AggregateProtocol Help


Copy link to this message
-
Re: AggregateProtocol Help
Gary Helmling 2012-01-02, 06:23
Hi Royston,

Try increasing the value set for hbase.rpc.timeout (by default 60
seconds).  Add something like this to hbase-site.xml:

<property>
<name>hbase.rpc.timeout</name>
<value>180000</value>
</property>

This would increase the timeout value to 3 minutes, for example.

But as Ted and Himanshu mention, 557k rows is not a large table and
should not be taking very long to process, unless these happen to be
very, very wide rows.  It sounds like there could be something else
going on.  Does debug level logging show anything else in the region
server log?

--gh

On Sun, Jan 1, 2012 at 5:53 PM, Ted Yu <[EMAIL PROTECTED]> wrote:
> Thanks for the reminder Himanshu.
>
> Royston:
> From this blog you can get some history on this subject:
> http://zhihongyu.blogspot.com/2011/03/genericizing-endpointcoprocessor.html
>
>
> On Sun, Jan 1, 2012 at 5:18 PM, Himanshu Vashishtha <[EMAIL PROTECTED]
>> wrote:
>
>> Hello Royston,
>>
>> Sorry to hear that you are getting trouble while using Aggregation
>> functionalities.
>>
>> 557k rows seems to be a small table and a SocketTimeout does not seem to be
>> an ok response.
>> It will be good to know the region distribution as such. (how many regions?
>> Is it a full table scan?)
>>
>> You are using the sum function; how are you using the ColumnInterpreter.
>> Can you enable the log level to debug to see why the RS is taking that long
>> to respond (more than 113 sec).
>> The 0 return value is the default result.
>>
>> Thanks for trying this out.
>>
>> Thanks,
>> Himanshu
>>
>> On Sun, Jan 1, 2012 at 12:26 PM, Royston Sellman <
>> [EMAIL PROTECTED]> wrote:
>>
>> > Hi Ted,
>> >
>> > I think 0 is the only value we ever see (I'll check tomorrow: the server
>> > is down right now). Our table has 557,000 rows. I'll try a much shorter
>> > table tomorrow.
>> >
>> > Yes, we have RS running on the NN, but it's a test cluster and we are
>> used
>> > to it :)
>> >
>> > Do you think using AggregationProtocol is the best strategy for the case
>> > where we want to use basic SQL-style functions like SUM, AVG, STD, MIN,
>> > MAX? Do you think there is a better strategy?
>> >
>> > Many thanks,
>> > Royston
>> >
>> >
>> > On 1 Jan 2012, at 17:58, Ted Yu wrote:
>> >
>> > > Royston:
>> > > Happy New Year to you too.
>> > >
>> > >>> java.net.SocketTimeoutException: Call to namenode/10.0.0.235:60020
>> failed
>> > on
>> > >
>> > > It seems the namenode above actually refers to a region server. This
>> is a
>> > > little bit confusing :-)
>> > >
>> > > The sum value below is 0.
>> > > Have you ever seen a value greater than 0 ?
>> > >
>> > > How many rows are there in this CF:CQ ?
>> > > The timeout was reported earlier by other people where there're many
>> rows
>> > > in the table.
>> > >
>> > > There is a JIRA to provide streaming support for coprocessor but the
>> > > development there has stalled.
>> > >
>> > > Cheers
>> > >
>> > > On Sun, Jan 1, 2012 at 9:35 AM, Royston Sellman <
>> > > [EMAIL PROTECTED]> wrote:
>> > >
>> > >> Hi Gary and Ted,
>> > >>
>> > >> Royston (Tom's colleague) here. Back onto this after the Christmas/New
>> > Year
>> > >> break.
>> > >>
>> > >> Many thanks for your help so far. We enabled our database via your
>> > >> hbase-site.xml mod and were able to move on. to other errors. But I
>> > think
>> > >> we
>> > >> are now actually getting an aggregation partially calculated on our
>> > table
>> > >> (this feels like progress). The details:
>> > >>
>> > >> On running our client we now get this exception:
>> > >> 11/12/31 17:51:09 WARN
>> > >> client.HConnectionManager$HConnectionImplementation: Error executing
>> for
>> > >> row
>> > >>
>> > >>  java.util.concurrent.ExecutionException:
>> > >> org.apache.hadoop.hbase.client.RetriesExhaustedException: Failed after
>> > >> attempts=10, exceptions:
>> > >> Sat Dec 31 17:41:30 GMT 2011,
>> > >> org.apache.hadoop.hbase.ipc.ExecRPCInvoker$1@1fc4f0f8,
>> > >> java.net.SocketTimeoutException: Call to namenode/10.0.0.235:60020