hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jinal Shah <jinalshah2...@gmail.com>
Subject Re: WARN messages clarifications
Date Thu, 02 Oct 2014 00:05:09 GMT
Just a follow-up on this. Does compaction add up to query execution time in
hbase?

On Wed, Oct 1, 2014 at 1:34 PM, Jinal Shah <jinalshah2007@gmail.com> wrote:

> Thanks Ted, that helps.
>
> On Wed, Oct 1, 2014 at 12:51 PM, Ted Yu <yuzhihong@gmail.com> wrote:
>
>> Suppose your OutlierHTable was communicating with a region server which
>> happened to go through GC pause.
>> The query would exhibit longer duration compared to one where no region
>> server goes through GC (assuming other factors being equal: network
>> traffic, size of data queried, etc).
>>
>> bq. Is GC pause set on a timer?
>>
>> I don't think so. It depends on the load region server handles.
>>
>> Cheers
>>
>> On Wed, Oct 1, 2014 at 10:37 AM, Jinal Shah <jinalshah2007@gmail.com>
>> wrote:
>>
>> > Is GC pause set on a timer? How much does that impact the query request?
>> >
>> > On Wed, Oct 1, 2014 at 10:48 AM, Ted Yu <yuzhihong@gmail.com> wrote:
>> >
>> > > One other possibility would be the region server going through GC
>> pause.
>> > >
>> > > Cheers
>> > >
>> > > On Wed, Oct 1, 2014 at 8:04 AM, Jinal Shah <jinalshah2007@gmail.com>
>> > > wrote:
>> > >
>> > > > We are using 0.94.15 version.
>> > > >
>> > > > On Wed, Oct 1, 2014 at 10:03 AM, Ted Yu <yuzhihong@gmail.com>
>> wrote:
>> > > >
>> > > > > I searched 0.94 and 0.98 code base but didn't find the message
>> below.
>> > > > >
>> > > > > Which version of HBase are you using ?
>> > > > >
>> > > > > Cheers
>> > > > >
>> > > > > On Wed, Oct 1, 2014 at 7:58 AM, Jinal Shah <
>> jinalshah2007@gmail.com>
>> > > > > wrote:
>> > > > >
>> > > > > > Hi everyone
>> > > > > >
>> > > > > > We are seeing these WARN log messages from this class
>> > > > > >
>> > > > >
>> > > >
>> > >
>> >
>> com.cerner.hbase.extensions.OutlierHTableInterceptorFactory$OutlierHTable.
>> > > > > > It normally happens when the query takes a long time to
execute
>> and
>> > > > > > displays the duration as well like this
>> > > > > >
>> > > > > > Duration Threshold exceeded: [durationThreshold=1000]
>> > > > > > Actuals: [actualBatchSize=1: actualPayloadSize=0:
>> > > actualDuration=3929]
>> > > > > >
>> > > > > > The scenarios I found for the delay are
>> > > > > >
>> > > > > >
>> > > > > >    - Network issues taking time for delays
>> > > > > >    - Trying to establish a connection for the first time
>> > > > > >    - Some component failures occurred like Zookeeper node
when
>> down
>> > > or
>> > > > > >    Master failed over.
>> > > > > >
>> > > > > >
>> > > > > > So can someone give say what other possible scenarios for
>> delays in
>> > > > > > executing a query.
>> > > > > >
>> > > > > > Thanks
>> > > > > > Jinal
>> > > > > >
>> > > > >
>> > > >
>> > >
>> >
>>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message