Excessive number of Oracle connections created from JDE...

swhitmire

Reputable Poster
So, it turns out my Vertex problem from yesterday wasn't just a Vertex problem. With just JDE running we are rapidly running out of database processes -- in 30 minutes from a restart we use up the 1500 that have been sufficient for the last four years. I have no idea what's causing this, it just started Monday, and all the changes we made have been put back. This is Oracle 11.2.0.3 on AIX 7.1, apps 9.0.2, tools 9.1.5.3. Any help is greatly appreciated... this is preventing work. I've opened a Sev 1 SR but haven't gotten help yet.
 
Some of the common things to check , that you may have already

When you group the count of sessions in v$session by user , machine and program where do you see the most processes coming from ? Kernel processes on logic server , JDBC connections from JAS , ube processes from batch server ? If any one stands out more than the other , that should give you a general direction to look in.

Did you make any changes around the F98OWSEC proxy users ( switched from common proxy user to individual user proxy accounts)



SELECT COUNT(*) , program ,machine,username FROM v$session GROUP BY program ,machine , username ORDER BY 1 DESC;

Based on your results you can then filter to the specific machine , program and list the process ID and then correlate that back to the E1 process ( check process ID in Server Manager )


Are your batch and app sever Oracle clients using shared connections or dedicated connections ( did you make any change around this , going from shared to dedicated will easily increase the number of sessions drastically)
 
In the [JDBj-CONNECTION POOL] of jdbj.ini, what is your poolGrowth setting? Also check the number of database connections in Server Manager.
 
I think we found the problem -- a change to a DSI script was apparently causing it to create a ton of connections for each operation and not clean them up. I found it by switching the system user just for the JDE user that DSI uses to a new one so we could distinguish the connections it was making. We rolled back the change and it appears to be behaving better... there are a lot fewer users for second/third shift, so I won't be positive until tomorrow morning. Thanks for the replies!!
 
Back
Top