delayed start of netweaver system

[Former Member]

Since yesterday our SapNetweaver based Enterprise Portal has a huge delayed start. Normally it takes about 10min until the system is up & running.

Now it takes about 2 hours.

System: NetWeaver 7.30 SP8 on HPUX

In the Log dev_server0 are only garbage collection or some CMS entries between State changed from 2 (Starting framework) to 10 (Starting apps) and State changed from 10 (Starting apps) to 3 (Running).

No other log in /usr/sap/SID/InstanceNr/work directory is written while the systems starts (except for std_server0.out and available.log)

The default respectively the application log in /usr/sap/SID/Nr/j2ee/cluster/server0/log are not really helpful.

So has anyone a idea or a hint where i can find a useful entry in a log?

Best regards

André

[Former Member · Accepted solution]

Ok, it looks like old Oracle statistics where the reason. Their planed through crontab everyday, but after executing them manually the system now starts in under 10min. strange...

[Former Member]

forget to tell: after 10min the jstart process 'disappears'. which means, in the normal startup phase the jstart process running with many cpu resources.

so it looks like the system is idle.

[Former Member]

Hi Andre,

While starting have you cleared all the semaphore and clear all the locks.

I hope it should help to make it faster.

Regards,

Deepanshu Sharma

[Former Member]

I switched to another clusternode of our HA-Cluster so this shouldn't be a problem.

[Former Member]

Ok,

So does that resolve your query.

Regards,

Deepanshu Sharma

[Former Member]

No it didn'n solve the problem.

andré

[Former Member]

Hello André

In most of the cases, the J2EE Engine spends most of the time in Starting Applications phase. IN order to diagnose the issue, you need to capture 4-5 thread dumps at an interval of 1 min each on all server nodes. Generally, most of the Application Threads are blocked by a thread. Once you find such thread, you need to diagnose in that direction. There can be long tunning DB calls which can block applications threads. You may refer below sap link for capturing and analysing thread dumps:

(JSTTSG)(Deploy)Thread Dump - Technology Troubleshooting Guide - SCN Wiki

I hope this will help you!!

Thanks,

Tapan