Re: [Sequoiadb-discuss] [Sequoia] Problems with sequoia in Loadtests

[Toni M. L. <ton...@gm...>]

I configured Jconsole and it seems we don´t have any problem with Memory.

After some investigation I think that we have the problem not in sequoia, we
have it in JBOSS... I will tell you....

Thanks a lot for the help,

Toni.

2009/6/2 Toni Menendez Lopez <ton...@gm...>

> Master I manage to put the SQL Monitoring, and now my surprise is even
> great...
>
>
> Look this a trace in MySQL :
> select user0_.ID as ID19_0_, user0_.SPROLE_FK as SPROLE2_19_0_,
> user0_.IAPROLE_FK as IAPROLE3_19_0_, user0_.COROLE_FK as COROLE4_19_0_,
> user0_.NAME as NAME19_0_, user0_.PASSWORD as PASSWORD19_0_, user0_.ROLETYPE
> as ROLETYPE19_0_, user0_.USER_FK as USER8_19_0_, user0_.EMAIL as EMAIL19_0_,
> user0_.STATUS as STATUS19_0_, user0_.TRUSTED as TRUSTED19_0_,
> user0_.VALIDATIONUSERID as VALIDAT12_19_0_, user0_.CMSCO as CMSCO19_0_ from
> cdpdb.user user0_ where user0_.ID=21 Jun  2, 2009 10:19:17.149134000
>  Jun  2, 2009 10:19:17.149185000
> select roleiap0_.ID as ID16_0_, roleiap0_.USER_FK as USER2_16_0_,
> roleiap0_.NAME as NAME16_0_ from cdpdb.roleiap roleiap0_ where
> roleiap0_.ID=1 Jun  2, 2009 10:19:17.149605000
> select itvplat0_.ID as ID9_0_, itvplat0_.NAME as NAME9_0_,
> itvplat0_.BROADCAST as BROADCAST9_0_, itvplat0_.AUTHCLASS as AUTHCLASS9_0_,
> itvplat0_.PUBLISHER as PUBLISHER9_0_ from cdpdb.itvplat itvplat0_ where
> itvplat0_.ID=4 Jun  2, 2009 10:19:17.149856000
>  Jun  2, 2009 10:19:17.149876000
>  Jun  2, 2009 10:19:17.149881000
>  Jun  2, 2009 10:19:17.167730000
>  Jun  2, 2009 10:19:17.187738000
>  Jun  2, 2009 10:19:17.188534000
>  Jun  2, 2009 10:19:17.188737000
>  Jun  2, 2009 10:19:17.188743000
>  Jun  2, 2009 10:19:17.189901000
>  Jun  2, 2009 10:19:17.189905000
> select itvplat0_.ID as ID9_0_, itvplat0_.NAME as NAME9_0_,
> itvplat0_.BROADCAST as BROADCAST9_0_, itvplat0_.AUTHCLASS as AUTHCLASS9_0_,
> itvplat0_.PUBLISHER as PUBLISHER9_0_ from cdpdb.itvplat itvplat0_ where
> itvplat0_.ID=4 Jun  2, 2009 10:19:18.030538000
>  Jun  2, 2009 10:19:18.030604000
>  Jun  2, 2009 10:19:18.030620000
> select itvplat0_.ID as ID9_0_, itvplat0_.NAME as NAME9_0_,
> itvplat0_.BROADCAST as BROADCAST9_0_, itvplat0_.AUTHCLASS as AUTHCLASS9_0_,
> itvplat0_.PUBLISHER as PUBLISHER9_0_ from cdpdb.itvplat itvplat0_ where
> itvplat0_.ID=4 Jun  2, 2009 10:19:18.030825000
>
> Which it looks that between 10:19:17.149856000 and 10:19:18.030538000 we
> have a gap that no selects are executed.... ( remember that LoadTests are
> beeing executed so it is impossible this), but when I try to see the results
> of SQLMonitoring in Sequoia we don´t have any "Max time (ms):" bigger that
> 50, so...
>
> So, it seems that we have any kind of problem related to transmission or
> queue or I don´t no what....
>
> Do you know is there is any chance to monitor traffic between hibernate (
> JBOSS)  and the  input of sequoia, I think that it is  in port 25322 (
> controller ? ).
>
> Anyway, now I am going to check Jconsole !
>
> Toni.
>
>
>
>
> 2009/6/2 Emmanuel Cecchet <ma...@fr...>
>
> Toni,
>>
>> The thing is that the problem only ocurrs from time to time, and when
>>> high load is sent,
>>>
>>> Is there any way to tune sequoia? Or monitor sequoia when high load is
>>> sent ?
>>>
>>> -which are good jvm parameters?any way to monitor jvm ?
>>>
>>>
>> I would recommend to use jconsole (
>> http://java.sun.com/developer/technicalArticles/J2SE/jconsole.html) that
>> is shipped since J2SE 5 to monitor your JVM and find out if you have any
>> memory problems.
>>
>>> -how request caché works?
>>>
>>>
>> Sequoia has 3 caches as described in sequoia.dtd:
>> "  MetadataCache: this cache improves the ResultSet creation time by
>> keeping
>>  the various field information with their metadata. It is strongly
>> encouraged
>>  to use this cache that reduces both cpu and memory usage.
>>  ParsingCache: allows to parse a request only once for all its executions.
>>  This reduces the cpu load on the controller.
>>  ResultCache: this cache keeps the results associated to a given request.
>>  Cache entries can be invalidated according to various policies. This
>> cache
>>  reduces the load on the database backends."
>>
>> I would not recommend to use the result cache if you have memory problems.
>>
>>
>>> -for last,how many queries per second can we achice ?
>>>
>>> The thing is our application works quite bad with high load, and we
>>> have to improve the Database access.
>>>
>>>
>> For reads, there is no real limit for queries per second as long as you
>> have enough concurrent queries and enough resources to execute them.
>> Everything depends on your system setup and your workload.
>>
>> Hope this helps,
>>
>> Emmanuel
>>
>> --
>> Emmanuel Cecchet
>> FTO @ Frog Thinker Open Source Development & Consulting
>> --
>> Web: http://www.frogthinker.org
>> email: ma...@fr...
>> Skype: emmanuel_cecchet
>>
>>
>