Oracle Listener Broken Pipe

[Jim Turner]

I saw Gumeni's post from January 8, 2001
http://forums.itrc.hp.com/cm/QuestionAnswer/1,1150,0x65b9854994d9d4118fef0090279cd0f9,00.html
which documents the exact same problem I'm having. There were a couple of replies with ideas, but unfortunately nothing definitive and no follow-up by Gumeni.

Here's the error text my DBA sent over:
16-AUG-2001 09:58:54 * (CONNECT_DATA=(SID=qeu)(GLOBAL_NAME=qeuhpux22.world)(CID=(PROGRAM=)(HOST=hpux29)(USER=dtc0653))) * (ADDRESS=(PROTOCOL=tcp)(HOST=17
2.20.12.125)(PORT=49342)) * establish * qeu * 12500
TNS-12500: TNS:listener failed to start a dedicated server process
TNS-12547: TNS:lost contact
TNS-12560: TNS:protocol adapter error
TNS-00517: Lost contact
HPUX Error: 32: Broken pipe

He said the listener just died. He was able to restart it without any problems, but our European developers waxed greatly unhappy in a biblical sense over the interruption.

The config:
Model: 9000/800/N4000-55
OS: B.11.00 U
RAM: 6291456 Kbytes
Swap: 4.0 GB (dev)
Pseudoswap: yes
SW: ORACLE 8.1.6

Thanks to all, and only three days until HP World. Woohoo!!!

Cheers,
Jim

[Sridhar Bhaskarla]

- Check all your Kernel parameters are within
the limit
- Check the resources like CPU, Mem, SWAP

If you don't have any problem there, you can ask them to relink the oracle executables. It fixed the problem before.

-Sri

[Jim Turner]

Sri,

Thanks for the tips. The kernel parms and resources are just fine. Relinking Oracle seems like a shotgun solution at this point. I'll let the DBA's work the Oracle tar first.

Thanks,
Jim

[Manuel P. Ron]

You are rigth. You must rebuild your executable.

Go to the $ORACLE_HOME/sqlplus/lib directory

Execute the following command:

make -f ins_sqlplus.mk install

Good luck!

[Manuel P. Ron]

Sorry, the above command must be this two:

make -f ins_network.mk install
make -f ins_rdbms.mk install

[ajax13]

Hi,
Check the CONNECT_TIMEOUT_LISTENER in listener.ora file, try increase by 5 or 10 Sec.
Since the problem happened to people in Europe, it could be a network issue, the network is slow. Note if you set this parameter to 0, the connection could wait forever. If you don't see this in listener.ora then it defaults to 10 Seconds. You also would need to restart the listener after doing that change.
This may not be the solution, this is something I would look at and eliminate from being an issue.

Thanks

[Jim Turner]

Manuel,
I appreciate your advice. My DBA's take care of that sort of thing, and I leave it to their discretion. Nobody has yet to explain *why* a relink would fix the problem. So far it just seems like a blind recommendation.

Samir,
I'll pass your suggestion along to my DBA's.

Thanks to you both,
Jim

[Sridhar Bhaskarla]

I have absolutely no idea why we should relink. We had similar problem. After struggling for two days, they relinked the executables and the problem went away.
Just shared with you my experience. It is something that oracle has to answer.

-Sri

[Manuel P. Ron · Accepted solution]

Jim. It is NOT a blind recommendation. It's the OFFICIAL solution of Oracle Corp. about your question.
Though you don't believe me, I'm a expert Oracle DBA with a lots of installations over HP-UX 11.x with 32/64 bits and I've suffered this problem before.
And, *WHY* it happens? I think it is not the correct question in a HP-UX forum because this problem happens over UNIX and must be response for our Oracle workmates.
I've attached the note 2064862.102 realesed by Oracle. You can see, it is not a BLIND recomendation

[Jim Turner]

Samir,
Thanks for the recommendation -- we will hold it in reserve as the problem has yet to repeat and everything is running fine.

Sridhar,
Thanks for the frank, honest answer. I rank Orafice right up there with Microsloth for corporate snobbiness. Ellison can take his "Don't modify our software to suit your business practices -- modify your business practices to suit our software" bullsh*t and stuff it up his wazoo.

Manuel,
I never said I *didn't* believe you, I just objected to the seemingly "shotgun" solution. Nor do I hold you responsible for an "OFFICIAL solution of Oracle Corp". I have no doubt that a relink would remedy a variety of Oracle problems much the same way that reloading Winblows fixes most any Microsloth problem. Finally, I am not questioning your expertise as a DBA. I don't know a great deal of what our DBA's do, nor do I want to know. My baseball hat says "Operating System" on the front. I work with our DBAs as needed but we largely do our own things. I have 65 boxes to keep me busy all on my own.

Folks, please consider this issue closed.

Thanks to all!
Jim

[Alexander M. Ermes]

Hi there.
Just a word from an old assembler worm.
The relink does adjust the code to the CPU instructions. The executables delivered from Oracle are just a general 'let us try and hope it works '. On a Cobol program you can check the object list to see what assembler instructions hide behind Cobol coding. I have not found yet, where i can find that from relinking Oracle executables.
Rgds
Alexander M. Ermes

[Wodisch]

Hello Jim,

being a little bit late on that thread, I will try to explain
anyway:

I guess on that systems some patches have been
installed, "PHNE_*" ones, I guess...

Well, everytime you patch something that the Oracle
codefiles are linked to, you do have to relink the Oracle
codefiles.

That' it.

Just my ?0.02,
Wodisch

[Jim Turner]

Alexander,
Ah, a man after my own heart. I cut my teeth with machine code and assembler. I can still cipher the 2's complement of an addr for a reverse branch :-)
I do understand what you are describing with the mechanics of the relink, though. When something "breaks", I always backtrack to determine what changed just beforehand. In this case, there wasn't a change that could be correlated with "worked before, doesn't work now". The issue is further complicated by the fact that it has thus far been a one-time failure. It ran fine before, and it has been running fine since. Yes, this would indicate a temporary resource problem, but there was no indication (glance, top, mwa, etc.) of such.

Wodisch,
I'm all too familiar with what you describe. Again, the scenario would usually be "load patches, Oracle broken, relink Oracle, Oracle working". In this instance, no patches were recently loaded, and the failure appears to have been a one-time event.

I appreciate everyone's help. Really I do. I'm afraid this will fall into the category of "things that make you go 'huh?'"

Best regards to all,
Jim