Vimal,
In my case, for a single client my queries take less than a second (sub-second 
performance is  on what we were shooting for).
But the same queries when run concurrently gives completely degraded 
performance. That is the reason I wrote that no-op test case which I have 
attached in the test HBASE-9410 bug. So many queries trying to execute on a 
region lead to some sort of lock contentions, is my guess which is causing the 
slow performance. 
My test case runs fine on 0.95.2 but on 0.94.x even for about 8 concurrent 
clients it is very high and I see these errors in the log.
You should also try running the test case. You will see bad performance even 
though the coprocessor function does not do anything.
 
Regards,
- kiru


Kiru Pakkirisamy | webcloudtech.wordpress.com


________________________________
 From: Vimal Jain <[email protected]>
To: Kiru Pakkirisamy <[email protected]> 
Cc: Himanshu Vashishtha <[email protected]>; "[email protected]" 
<[email protected]>; lars hofhansl <[email protected]> 
Sent: Wednesday, October 23, 2013 10:50 PM
Subject: Re: Warning messages in hbase logs
 


Hi Kiru,
Can you please elaborate on this ?




On Mon, Oct 21, 2013 at 5:30 PM, Vimal Jain <[email protected]> wrote:

Hi Kiru,
>Thanks for the reply.
>My understanding is not very clear  with respect to these warning messages in 
>region server logs.
>When i read documentation , it says its related to slow query logs.
>So is it a issue in earlier versions of hbase and is fixed in 0.96 or is it an 
>issue in my configuration and i have to set some parameters to prevent this?
>Please help in understanding the exact issue.
>
>
>
>
>On Sat, Oct 19, 2013 at 3:31 AM, Kiru Pakkirisamy <[email protected]> 
>wrote:
>
>Vimal,
>>Maybe you should try running the test case for HBASE-9410 (with some 
>>modifications) and report back.
>>This way Lars/committers/contributors can work on this bug with confirmation.
>>
>>
>>
>>The test case makes a coprocessor call (which does nothing) on a table with 
>>regions in all the region servers.
>>Feel free to send me any email in this regards. 
>> 
>>Regards,
>>- kiru
>>
>>
>>
>>Kiru Pakkirisamy | webcloudtech.wordpress.com
>>
>>
>>________________________________
>> From: Vimal Jain <[email protected]>
>>To: Himanshu Vashishtha <[email protected]> 
>>Cc: "[email protected]" <[email protected]>; Kiru Pakkirisamy 
>><[email protected]>; lars hofhansl <[email protected]> 
>>Sent: Thursday, October 17, 2013 9:39 PM
>>
>>Subject: Re: Warning messages in hbase logs
>> 
>>
>>
>>Hi.
>>I am still getting these warnings in region server logs.
>>I am still using hbase 0.94.7.
>>Is there any impact of these warning messages ?
>>
>>
>>
>>
>>On Thu, Sep 26, 2013 at 11:19 PM, Himanshu Vashishtha <[email protected]> 
>>wrote:
>>
>>
>>>
>>>
>>>
>>>
>>>On Thu, Sep 26, 2013 at 4:07 AM, Kiru Pakkirisamy 
>>><[email protected]> wrote:
>>>
>>>I was on 0.94.11 and still got the same msgs as Vimal. 
>>>>Ultimately, I moved to 0.95.2.
>>>>
>>> 
>>>You know it is a dev release, right? 
>>>
>>>
>>>
>>>
>>>Regards,
>>>>- kiru
>>>>
>>>>
>>>>Kiru Pakkirisamy | webcloudtech.wordpress.com
>>>>
>>>>
>>>>
>>>>________________________________
>>>> From: lars hofhansl <[email protected]>
>>>>To: Vimal Jain <[email protected]>; "[email protected]" 
>>>><[email protected]>
>>>>Cc: Kiru Pakkirisamy <[email protected]>
>>>>Sent: Wednesday, September 25, 2013 9:46 PM
>>>>
>>>>Subject: Re: Warning messages in hbase logs
>>>>
>>>>
>>>>You probably ran into
>>>>[HBASE-6870] - HTable#coprocessorExec always scan the whole table (0.94.8)
>>>>
>>>>Or maybe one of these:
>>>>
>>>>[HBASE-8806] - Row locks are acquired repeatedly in 
>>>>HRegion.doMiniBatchMutation for duplicate rows. (0.94.10)
>>>>[HBASE-8639] - Poor performance of htable#getscanner in multithreaded 
>>>>environment due to DNS.getDefaultHost() being called in 
>>>>ScannerCallable#prepare() (0.94.9)
>>>>[HBASE-9087] - Handlers being blocked during reads (0.94.11)
>>>>
>>>>
>>>>-- Lars
>>>>
>>>>
>>>>
>>>>________________________________
>>>>From: Vimal Jain <[email protected]>
>>>>To: [email protected]; lars hofhansl <[email protected]>
>>>>Cc: Kiru Pakkirisamy <[email protected]>
>>>>Sent: Wednesday, September 25, 2013 10:01 PM
>>>>Subject: Re: Warning messages in hbase logs
>>>>
>>>>
>>>>
>>>>I am using  Hadoop version - 1.1.2 , Hbase version - 0.94.7 .
>>>>
>>>>
>>>>
>>>>On Thu, Sep 26, 2013 at 9:38 AM, lars hofhansl <[email protected]> wrote:
>>>>
>>>>Eh? That should not be the case, we should figure where 0.95.2 is different 
>>>>from 0.94.
>>>>>With which version of the 0.94 did you see this?
>>>>>
>>>>>-- Lars
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>----- Original Message -----
>>>>>From: Kiru Pakkirisamy <[email protected]>
>>>>>To: "[email protected]" <[email protected]>
>>>>>Cc:
>>>>>Sent: Tuesday, September 24, 2013 9:46 PM
>>>>>Subject: Re: Warning messages in hbase logs
>>>>>
>>>>>Are you running many concurrent clients ? I had a similar problem when 
>>>>>running on 0.94.x and I moved to 0.95.2 for this reason (see HBASE-9410)
>>>>> 
>>>>>Regards,
>>>>>- kiru
>>>>>
>>>>>
>>>>>________________________________
>>>>>From: Vimal Jain <[email protected]>
>>>>>To: [email protected]
>>>>>Sent: Tuesday, September 24, 2013 9:36 PM
>>>>>Subject: Warning messages in hbase logs
>>>>>
>>>>>
>>>>>Hi,
>>>>>I am seeing some warning message in Hregion log file  for quite a few days.
>>>>>The message is :
>>>>>
>>>>>*WARN org.apache.hadoop.ipc.HBaseServer: (responseTooSlow):
>>>>>{"processingtimems":12994,"call":"multi(org.apache.hadoop.hbase.client.MultiAction@602cdaf7),
>>>>>rpc version=1, client version=29, 
>>>>>methodsFingerPrint=-1368823753","client":"
>>>>>192.168.20.31:42206
>>>>>","starttimems":1380057021850,"queuetimems":0,"class":"HRegionServer","responsesize":0,"method":"multi"}
>>>>>*
>>>>>
>>>>>
>>>>>Can someone please help in understanding what this warning message is ?
>>>>>
>>>>>--
>>>>>Thanks and Regards,
>>>>>Vimal Jain
>>>>>
>>>>
>>>>
>>>>--
>>>>
>>>>Thanks and Regards,
>>>>Vimal Jain
>>>
>>
>>
>>-- 
>>
>>Thanks and Regards,
>>Vimal Jain
>>
>>
>
>
>-- 
>
>Thanks and Regards,
>Vimal Jain


-- 

Thanks and Regards,
Vimal Jain

Reply via email to