Scheduler Task hangs portal
I am seeing an interesting quirk in the portal and am wondering whether anyone else has ever experienced this.
I have a NetWeaver 7.0 Enterprise Portal running on AIX 5.3 and Oracle 10.2.0.2.
After the system was live for about 6-8 weeks, the database got hung up because the default number of Oracle processes (80) was exceeded.
The value was increased to 1024 and about 6-8 weeks later the database connections exceeded that value. This symptom repeated again after the Oracle parameter 'processes' was increased to 2048.
According to the notes for tuning the Oracle 10G database, the portal should not exceed 280 processes.
The timing usually falls late at night, which lead me to the Scheduler Tasks, since there is no other user activity taking place at the time. Typically the 1000+ database connections fall within a short timespan and will continue until the issue runs its course (24-30 hours) or until the system is restarted.
I moved the Scheduler tasks off of the central instance and onto an application server. Several days later, outside of business hours, the server0 jlaunch process hung up on the application server and at the same time 1000+ database connections were made on the central instance.
The open connections between the application server and the central instance occurred with the following frequency:
7:15AM, 5:45PM, 8:00PM, 8:00:03PM, 8:15PM, 8:15PM, 8:30PM, 8:30PM, 8:45PM, 8:45PM, 8:55:02PM, 9:00PM, 9:00PM, 9:00PM, 9:00:02PM, 9:00:03PM, 9:00:03PM, 9:03:45PM, 9:05PM, 9:05:45PM, 9:09:08PM, 9:10PM.... and so on with 42 more connections from 9:10 to 9:59 and 1925 connections from 10:00 to 10:59. The initial frequency of the every 15 minutes or so brings my focus to the GC Scheduler Tasks (CM Repository GC & Web Repository GC).
I am looking for ideas on how to troubleshoot this bug further, and validation as to whether anyone else is experiencing a similar issue.