URGENT: svc_tli_create: other side wants torelease connection ERROR

[Sanjay Verma]

Hi Friends,

When ever I want to make a connection to the HP-UX 11.00 server, the connection fails and the following error mesg. is logged in the syslog.file:

svc_tli_create: other side wants torelease connection

This problem only occurs for a specific time only (approx 30mins) between 8:00am - 9:00am.

I'll really appreciate for your suggestions / solutions on this.

Thanks & Regards,
Sanjay

[Michael Tully]

Hi,

Have a look at this posting. I am sure it has to do with some sort of remote login. The suggestion here is that patches may need to be loaded.

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x266c37f45ef7d4118fef0090279cd0f9,00.html

HTH
~Michael~

[Sanjay Verma]

Hi Michael,

Thanks for your quick response. Seems all the given patches are installed in the server.

But still getting the same error mesg. but only for 30mins or so in a day.

Regards,
Sanjay

[Michael Tully]

Hi,

For that period of the day, perhaps you should consider logging various bits of traffic, telnet, ftp, rlogin.

Michael

[Sanjay Verma]

Hi Michael,

The performance and everything is normal.
Users connect normally and for few mins. all the new connection only gets refused and after few mins., everthing is back to normal.

In the syslog file, there's no entry of svc_tli_create apart from that specific period.

Amazing isn't it? Anyway, researching more into this.

Regards,
Sanjay

[Michael Tully]

Hi,

In my first response, I referred to patching, are these up to date?

Michael

[Brian Hackley]

Sanjay,
The message sounds like "SOMETHING in the middle" between the users and your program is causing the RPC connection to be blocked for some odd reason or other. I would say that you MIGHT get more data from the HP-UX side of the connection with a nettl trace. You'll need to know the IP address of the client trying the connection. I don't know if that is logged anywhere by your program.

More likely getting a network sniffer to snoop around the network, will help you isolate what's really doing this, better than a host-based trace.
The last thing I can say is that since this appears to be time-dependent, check out crontabs to if anything being started up to cause the issue.

hope this helps,

-> Brian Hackley

[Sanjay Verma]

Hi Brian,

Thanks for your comments on this. Started the nettl process but it generates the error mesg. in binary which can be seen with the strings command.

But still not able to get enough information.

Will diagnose further..

Thanks
Sanjay

[T G Manikandan]

nettl stores information in binary format
YOu have to use netfmt -f to change in readable format from binary.
see netfmt man pages

Thanks

[Sanjay Verma]

Hi T.G.,

Thanks for providing the valuable information. Now, should I require to change the output file from the default /var/adm/nettl* to anyother file, it captures on the old file itelf. Any mistake in the nettl command?

Regards,
Sanjay

[T G Manikandan]

do this

1) start trace putting data into 1MB trace file ( /tmp/raw.TRC0, raw.TRC1).
/etc/nettl -tn -pduin -pduout -e all -f /tmp/raw

2) stop trace -> /etc/nettl -tf -e all
3) Format traces into a file like this :

/etc/netfmt -N -n -l -f /tmp/raw.TRC0 > /tmp/file0

/etc/netfmt -N -n -l -f /tmp/raw.TRC1 > /tmp/file1


-n option -> print IP addresses, not hostnames

[Sanjay Verma]

Hi Friends,

Thanks for all of your efforts and time taken to give possible solutions. Now, this issue has been resolved and the following issues have been identified as the root cause for this:

(a) The DNS setup of the server was causing this issue.
(b) Added unresolved IP Addresses to /etc/hosts file. This means, the server will not use the DNS to seek unresolved IP addresses.

Thanks once again.

Cheers,
Sanjay