tunig tcp/ip or kernel parameters

[Shivkumar]

Hi,

We are using couple of applications on hpux 11i. Sometime we observe one application gets re-initialized. We get the alert but whenever click on the link it works fine.

I suspect we need to tune some tcp/ip or kernel parameters.

Can anyone suggest which tcp/ip or kernel parameters will fix this ?

Thanks,
Shiv

[Steven E. Protter]

Shalom Shiv,

tcp parameters are controlled by the ndd utility.

http://devrsrc1.external.hp.com/STK/impacts/i692.html

That link should be helpful.

SEP

[Bill Hassell]

Changing tcp/ip or kernel values without knowing the cause is just a waste of time. You need to trace the application through logfiles, webserver logs, syslog.log and perhaps execution traces. There is simply no useful information from the description you have given.

[Darrel Louis]

Shivkumar,

What is the application using?
Websphere, Oracle, Java etc.

What's the alert you're receiving?

Check the following url for info regarding Websphere.
http://publib.boulder.ibm.com/infocenter/wasinfo/v5r0/index.jsp?topic=/com.ibm.websphere.base.doc/info/aes/ae/tprf_tuneopsys.html

Darrel

[Shivkumar]

It is running the applications WebLogic7.0 and SiteMinder6.0. SiteMinder service gets re-initiazed very often. The process running SiteMinder goes down and comes back automatically. We don't see anything in various logs.

Need to know which tcp/ip or kernel parameter will fix the issue.

Regards,
Shiv

[Michael Steele_2]

As your cohorts have suggested gather more information. Especially if this is a production machine. However, if a lab or dev. machine then look into these kernel parameters as min. vals.

â Maxdsize 0x79000000
â Maxssize 0x8000000
â Maxtsize 0x8000000
â Maxfiles 1024
â Maxusers 64 (to indirectly boost NKTHREAD, used below)
â Max_thread_proc (NKTHREAD-10)

[Shivkumar]

Appreciate more suggestion and guidance if gurus can provide.

Regards,
Shiv

[James R. Ferguson]

Hi Shiv:

I second Bill's advice. In the absence of any error logs or specific standard 'errno's you have no idea which kernel (or network) parameter to change!

Does your application appear to leak memory? Does the failure occur only under load as for instance when n-number of connections are achieved? Have you run 'sar' and 'glance' and looked at the high-water marks for key kernel tables?

Consider, Shiv, what parameters you would change? You could make matters worse or better or you would see no change, and equally importantly, *why*?. Would you attempt to change more than one paramter at a time? What if that meant a reboot for every change? After changing a value, how would you measure the difference?

You simply have too little information.

Regards!

...JRF...

[Arunvijai_4]

Hi Shiv,

As HP-UX gurus mentioned, you should specify the apps you are using on 11i. Different apps have different settings. Say for example, Oracle and Weblogic.

Please provide some more information.

-Arun

[Shivkumar]

Arun, We are using WebLogic7.0 and SiteMinder6.0.

We didn't find any errors for applications in any of the log.

Just SiteMinder6.0 service gets re-initialized.

Thanks,
Shiv

[James R. Ferguson]

Hi SHiv:

In another thread of yours you had asked for information about Weblogic:

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1029821

Regards!

...JRF...

[Shivkumar]

Yes James. We have different Servers. Some running only weblogic and some running weblogic and siteminder both.

Regards,
Shiv

[Chan 007 · Accepted solution]

Shiv,

Please find the ndd (ftp) site access. It is very useful.

ftp://ftp.cup.hp.com/dist/networking/briefs/annotated_ndd.txt

Why can't you post you sm.log (Siteminder Log), Can you check any for the following.

1. Any specific (Day or time)
2. Any specific application
3. Check for 3rd party penetration testing. This may happen from your organisation, which will try to access your SM server. (Check syslog.log)
4. Check SM (Guide)

Chan

[rick jones]

At the risk of asking a silly question, what precisely do you mean by "reinitialized" here? Processes dieing and being restarted?

[Shivkumar]

Rick,

You are right. The Processes were dieing and being restarted.

I don't know whether it was a silly or intelligent question.

Regards,
Shiv

[rick jones]

Thanks - In broad handwaving terms, networking settings are unlikely to affect processes dying. Now, if you have processes that toss-up a "connection aborted" message before they die perhaps networking is involved, but otherwise, checking the kernel tunables to make sure they are sized for the app, and perhaps trying to take some tusc traces to try to catch a dying process would be useful. That, and looking for core files.

[Shivkumar]

Thanks Rick!!

warm regards.
Shiv