Hi Chalitha,
We checked this with local Oracle (Oracle express edition) setup. Following
are the investigations.
1. Initially there are only two inactive connections.
SQL> SELECT SID, SERIAL#, STATUS, SERVER FROM V$SESSION WHERE
USERNAME='GREGREGISTRY';
SID SERIAL# STATUS SERVER
---------- ---------- -------- ---------
16 51 INACTIVE DEDICATED
103 53 INACTIVE DEDICATED
2. Import a registry and add some resources - At thus point the number of
connections increase and there are both active and inactive connections.
This is possibly due to indexing thread continuously access the REG_LOG
table and it depends on the number of indexing threads.
3. After some idle time, the number of connections drop down to above two
inactive connections.
4. Go into overview page and view the asset details. This would probably
increase the connections if it doesn't get served from cache.
5. After some idle time it again drops down to two inactive connections
I suppose there are some settings to be done on the Oracle end to terminate
connections. Is this getting reproduced in MySQL, MariaDB in your end?
Shazni Nazeer
Mob : +94 777737331
LinkedIn : http://lk.linkedin.com/in/shazninazeer
Blog : http://shazninazeer.blogspot.com
On Thu, Aug 20, 2015 at 10:15 AM, Chalitha Waldeniyage <chali...@wso2.com>
wrote:
> Hi Chandana,
>
> We have reproduced this with clean databases as well. server start-up
> seems like several registry artifacts trying to establish db connection and
> connection pool exhausted and getting the issue reported in [1]. It's
> continuously growing the active connections.In order to recover database
> server needed to be restart.
>
> [1] https://wso2.org/jira/browse/REGISTRY-2762
>
>
>
> On Fri, Aug 14, 2015 at 12:26 PM, Chalitha Waldeniyage <chali...@wso2.com>
> wrote:
>
>> As per the load test on Greg 5.0.0 alpha,we have run a Jmeter script on
>> the QA two instances.
>> This jmeter script initially copying a text file on the registry root
>> location . This works fine and we have worked with the setup after the load
>> test. But again we have put a load on the
>> /_system/governance/publisher_notes/ location copying 7000 artifacts.
>> After this load test while restarting the server we observed that the
>> startup hangs at org.wso2.carbon.core.init.CarbonServerManager,
>> repository/deployment/server/
>>
>> We enabled the log4j.logger.org.wso2 DEBUG logs and noticed that it's
>> running a CacheCleanupTask and indexing the connection.
>>
>> According to the offline chat with the team, our first approach was
>> strategically wrong which was coping the 10,000 artifacts to the root
>> location. Therefore we have worked on following two workarounds
>>
>> [1] In order to clear the registry H2 space, remove H2 database from
>> repository/databases.
>>
>> [2] Start with the fresh pack but pointing to the same old gov and conf
>> dabases.
>>
>> But any of these workarounds suggested did not solve the original issue
>> we are facing (server hanging when starting the server)
>>
>> Right now we are trying to reproduce the scenario again by dropping the
>> gov/confg dbs & running the load test again under _system/governance/. We
>> will update the threads with the findings.
>>
>>
>> Thank you,
>> Chalitha.
>> --
>> *Chalitha Maheshwari*
>> Software Engineer-QA,
>> WSO2 Inc.
>>
>> *E-mail:* chali...@wso2.com
>> *Mobile: *+94710 411 112
>>
>
>
>
> --
> *Chalitha Maheshwari*
> Software Engineer-QA,
> WSO2 Inc.
>
> *E-mail:* chali...@wso2.com
> *Mobile: *+94710 411 112
>
_______________________________________________
Dev mailing list
Dev@wso2.org
http://wso2.org/cgi-bin/mailman/listinfo/dev
