You are viewing a plain text version of this content. The canonical link for it is here.
Posted to solr-user@lucene.apache.org by Marc Sturlese <ma...@gmail.com> on 2009/02/20 18:13:59 UTC
Re: concurrency problem with delta-import (indexing various cores
simultaniously)
Hey,
Yeah, I patched the bug reported by Ryuuichi of the SimpleDateFormat aswell.
Is there any other known concurrency bug that maybe I am missing?
In my use case I could manage to index not concurrently but would like to
discover why this is happening...
Thank you very much!
Shalin Shekhar Mangar wrote:
>
> On Fri, Feb 20, 2009 at 8:41 PM, Marc Sturlese
> <ma...@gmail.com>wrote:
>
>>
>> Hey there,
>> I am indexing 3 cores concurrently from 3 diferent mysql tables (I do it
>> every 5 minutes with a cron job).
>> The three cores use JdbcDataSource as datasource in data-config.xml
>> Reached a point, the core that fetches more mysql rows starts running so
>> so
>> solw until the thread seems to stop (but the other tow keep working
>> fine)...but java doesn't throw and exception...
>> I am using a nightly from early january. I found someone experienced the
>> same problem and uploaded a templateString patch to make it thread-save.
>>
>
> Marc, I'd strongly recommend using a more recent nightly build. There was
> another problem related to unsafe usage of SimpleDateFormat which was
> fixed
> recently.
>
> See https://issues.apache.org/jira/browse/SOLR-1017 (which was fixed on
> 11th
> Feb)
> --
> Regards,
> Shalin Shekhar Mangar.
>
>
--
View this message in context: http://www.nabble.com/concurrency-problem-with-delta-import-%28indexing-various-cores-simultaniously%29-tp22120430p22123287.html
Sent from the Solr - User mailing list archive at Nabble.com.
Re: concurrency problem with delta-import (indexing various cores
simultaniously)
Posted by Marc Sturlese <ma...@gmail.com>.
Thank you so much Ryuuichi, I was completelly stuck with this problem!
Ryuuichi KUMAI wrote:
>
> Hello Marc,
>
> I faced the similar problem, and I found a workaround.
> If the performance degradation in your application is caused by GC,
> this information might help you:
>
> https://issues.apache.org/jira/browse/SOLR-1042
>
> Regards,
> Ryuuichi Kumai.
>
> 2009/2/21 Marc Sturlese <ma...@gmail.com>:
>>
>> I am working with 3 index of 1 gig each. I am using the standard setting
>> of
>> the GC, haven't changed anything and using java version "1.6.0_07".
>> I don't know so much about GV configuration... just read this
>>
>> http://marcus.net/blog/2007/11/10/solr-search-and-java-gc-tuning/
>>
>> when a month ago I exeprienced another problem with Solr (at the end it
>> was
>> not GV's fault). So, any advice about wich GC should I try or what should
>> I
>> tune?
>>
>> Thank you very much!
>>
>>
>>
>> Shalin Shekhar Mangar wrote:
>>>
>>> On Fri, Feb 20, 2009 at 11:23 PM, Marc Sturlese
>>> <ma...@gmail.com>wrote:
>>>
>>>>
>>>> Yes,
>>>> Now it's almost tree days non-stop since I am running updates with the
>>>> 3
>>>> cores with cron jobs. If there are updates of 10000 docs everything is
>>>> alrite. When I start doing updates of 300000 is when that core runs
>>>> really
>>>> slow. I have to abort the import in that core and keep updating with
>>>> less
>>>> rows each time.
>>>> Another thing to point is that tomcat reaches the maximum memory I
>>>> allow
>>>> (2Gig) and never goes down (but at least it doesn't run out of memory).
>>>> Is
>>>> that normal? Shouldn't the memory go down a lot after an update is
>>>> completed?
>>>>
>>>
>>> I guess you are being hit by garbage collection. Memory utilization
>>> should
>>> go down once an import completes. Which GC are you using? There have
>>> been
>>> a
>>> few recent threads on GC settings. Perhaps you can try out a few of
>>> those
>>> settings. I don't know how big your documents/index are but if possible
>>> give
>>> it more memory.
>>>
>>> --
>>> Regards,
>>> Shalin Shekhar Mangar.
>>>
>>>
>>
>> --
>> View this message in context:
>> http://www.nabble.com/concurrency-problem-with-delta-import-%28indexing-various-cores-simultaniously%29-tp22120430p22125716.html
>> Sent from the Solr - User mailing list archive at Nabble.com.
>>
>>
>
>
--
View this message in context: http://www.nabble.com/concurrency-problem-with-delta-import-%28indexing-various-cores-simultaniously%29-tp22120430p22265652.html
Sent from the Solr - User mailing list archive at Nabble.com.
Re: concurrency problem with delta-import (indexing various cores
simultaniously)
Posted by Ryuuichi KUMAI <ry...@gmail.com>.
Hello Marc,
I faced the similar problem, and I found a workaround.
If the performance degradation in your application is caused by GC,
this information might help you:
https://issues.apache.org/jira/browse/SOLR-1042
Regards,
Ryuuichi Kumai.
2009/2/21 Marc Sturlese <ma...@gmail.com>:
>
> I am working with 3 index of 1 gig each. I am using the standard setting of
> the GC, haven't changed anything and using java version "1.6.0_07".
> I don't know so much about GV configuration... just read this
>
> http://marcus.net/blog/2007/11/10/solr-search-and-java-gc-tuning/
>
> when a month ago I exeprienced another problem with Solr (at the end it was
> not GV's fault). So, any advice about wich GC should I try or what should I
> tune?
>
> Thank you very much!
>
>
>
> Shalin Shekhar Mangar wrote:
>>
>> On Fri, Feb 20, 2009 at 11:23 PM, Marc Sturlese
>> <ma...@gmail.com>wrote:
>>
>>>
>>> Yes,
>>> Now it's almost tree days non-stop since I am running updates with the 3
>>> cores with cron jobs. If there are updates of 10000 docs everything is
>>> alrite. When I start doing updates of 300000 is when that core runs
>>> really
>>> slow. I have to abort the import in that core and keep updating with less
>>> rows each time.
>>> Another thing to point is that tomcat reaches the maximum memory I allow
>>> (2Gig) and never goes down (but at least it doesn't run out of memory).
>>> Is
>>> that normal? Shouldn't the memory go down a lot after an update is
>>> completed?
>>>
>>
>> I guess you are being hit by garbage collection. Memory utilization should
>> go down once an import completes. Which GC are you using? There have been
>> a
>> few recent threads on GC settings. Perhaps you can try out a few of those
>> settings. I don't know how big your documents/index are but if possible
>> give
>> it more memory.
>>
>> --
>> Regards,
>> Shalin Shekhar Mangar.
>>
>>
>
> --
> View this message in context: http://www.nabble.com/concurrency-problem-with-delta-import-%28indexing-various-cores-simultaniously%29-tp22120430p22125716.html
> Sent from the Solr - User mailing list archive at Nabble.com.
>
>
Re: concurrency problem with delta-import (indexing various cores
simultaniously)
Posted by Marc Sturlese <ma...@gmail.com>.
I am working with 3 index of 1 gig each. I am using the standard setting of
the GC, haven't changed anything and using java version "1.6.0_07".
I don't know so much about GV configuration... just read this
http://marcus.net/blog/2007/11/10/solr-search-and-java-gc-tuning/
when a month ago I exeprienced another problem with Solr (at the end it was
not GV's fault). So, any advice about wich GC should I try or what should I
tune?
Thank you very much!
Shalin Shekhar Mangar wrote:
>
> On Fri, Feb 20, 2009 at 11:23 PM, Marc Sturlese
> <ma...@gmail.com>wrote:
>
>>
>> Yes,
>> Now it's almost tree days non-stop since I am running updates with the 3
>> cores with cron jobs. If there are updates of 10000 docs everything is
>> alrite. When I start doing updates of 300000 is when that core runs
>> really
>> slow. I have to abort the import in that core and keep updating with less
>> rows each time.
>> Another thing to point is that tomcat reaches the maximum memory I allow
>> (2Gig) and never goes down (but at least it doesn't run out of memory).
>> Is
>> that normal? Shouldn't the memory go down a lot after an update is
>> completed?
>>
>
> I guess you are being hit by garbage collection. Memory utilization should
> go down once an import completes. Which GC are you using? There have been
> a
> few recent threads on GC settings. Perhaps you can try out a few of those
> settings. I don't know how big your documents/index are but if possible
> give
> it more memory.
>
> --
> Regards,
> Shalin Shekhar Mangar.
>
>
--
View this message in context: http://www.nabble.com/concurrency-problem-with-delta-import-%28indexing-various-cores-simultaniously%29-tp22120430p22125716.html
Sent from the Solr - User mailing list archive at Nabble.com.
Re: concurrency problem with delta-import (indexing various cores
simultaniously)
Posted by Shalin Shekhar Mangar <sh...@gmail.com>.
On Fri, Feb 20, 2009 at 11:23 PM, Marc Sturlese <ma...@gmail.com>wrote:
>
> Yes,
> Now it's almost tree days non-stop since I am running updates with the 3
> cores with cron jobs. If there are updates of 10000 docs everything is
> alrite. When I start doing updates of 300000 is when that core runs really
> slow. I have to abort the import in that core and keep updating with less
> rows each time.
> Another thing to point is that tomcat reaches the maximum memory I allow
> (2Gig) and never goes down (but at least it doesn't run out of memory). Is
> that normal? Shouldn't the memory go down a lot after an update is
> completed?
>
I guess you are being hit by garbage collection. Memory utilization should
go down once an import completes. Which GC are you using? There have been a
few recent threads on GC settings. Perhaps you can try out a few of those
settings. I don't know how big your documents/index are but if possible give
it more memory.
--
Regards,
Shalin Shekhar Mangar.
Re: concurrency problem with delta-import (indexing various cores
simultaniously)
Posted by Marc Sturlese <ma...@gmail.com>.
Yes,
Now it's almost tree days non-stop since I am running updates with the 3
cores with cron jobs. If there are updates of 10000 docs everything is
alrite. When I start doing updates of 300000 is when that core runs really
slow. I have to abort the import in that core and keep updating with less
rows each time.
Another thing to point is that tomcat reaches the maximum memory I allow
(2Gig) and never goes down (but at least it doesn't run out of memory). Is
that normal? Shouldn't the memory go down a lot after an update is
completed?
Thank you very much!
Shalin Shekhar Mangar wrote:
>
> On Fri, Feb 20, 2009 at 10:43 PM, Marc Sturlese
> <ma...@gmail.com>wrote:
>
>>
>> Hey,
>> Yeah, I patched the bug reported by Ryuuichi of the SimpleDateFormat
>> aswell.
>> Is there any other known concurrency bug that maybe I am missing?
>> In my use case I could manage to index not concurrently but would like to
>> discover why this is happening...
>>
>> Thank you very much!
>>
>>
> I don't see any obvious issue except for these two fixes. Are you
> experiencing this problem even after applying both of Ryuuichi's fixes?
>
> --
> Regards,
> Shalin Shekhar Mangar.
>
>
--
View this message in context: http://www.nabble.com/concurrency-problem-with-delta-import-%28indexing-various-cores-simultaniously%29-tp22120430p22125443.html
Sent from the Solr - User mailing list archive at Nabble.com.
Re: concurrency problem with delta-import (indexing various cores
simultaniously)
Posted by Shalin Shekhar Mangar <sh...@gmail.com>.
On Fri, Feb 20, 2009 at 10:43 PM, Marc Sturlese <ma...@gmail.com>wrote:
>
> Hey,
> Yeah, I patched the bug reported by Ryuuichi of the SimpleDateFormat
> aswell.
> Is there any other known concurrency bug that maybe I am missing?
> In my use case I could manage to index not concurrently but would like to
> discover why this is happening...
>
> Thank you very much!
>
>
I don't see any obvious issue except for these two fixes. Are you
experiencing this problem even after applying both of Ryuuichi's fixes?
--
Regards,
Shalin Shekhar Mangar.