[omniORB] Name service problems

[Roger wenham]

Hi,

I am having problems with omniNames stopping for no apparent reason.

This happens about once a week, on various Solaris machines.

Has anyone else seen this behaviour, if so, I would be grateful for any pointers
on how to go about debugging this.

The start command looks like this
/opt/omniORB/bin/omniNames -errlog /opt/log/omniName.log

There are no errors to be found in the omniName.log.

A second question is: When I re-start the name server after an error, it has of course
forgotten all the names registered with it. I have searched the documentation for the
way to make the name server non volatile, but have found nothing so far.

Can some kind soul enlighten me please...

Thanks in advance

Roger

[Michael]

omninames is non-valatile by default. It normally stores information in
/var/log/omninames-<hostname>.[log|bak] which needs to be writable by
the user running the process.

You can override this location by putting OMNINAMES_LOGDIR in the
environment or by specifying -logdir <directory> on the command line.

We're running omniNames out of daemontools (http://cr.yp.to) to make
sure it restarts on crashes - but we've yet to see one of these :)

Post by Roger wenham
Hi,
I am having problems with omniNames stopping for no apparent reason.
This happens about once a week, on various Solaris machines.
Has anyone else seen this behaviour, if so, I would be grateful for any pointers
on how to go about debugging this.
The start command looks like this
/opt/omniORB/bin/omniNames -errlog /opt/log/omniName.log
There are no errors to be found in the omniName.log.
A second question is: When I re-start the name server after an error, it has of course
forgotten all the names registered with it. I have searched the documentation for the
way to make the name server non volatile, but have found nothing so far.
Can some kind soul enlighten me please...
Thanks in advance
Roger

[Michael]

Hmmm I can't see anything in here preventing this from working. Maybe
it's because you're running on a non-standard port? Did you specify a
distinct endpoint when starting omniNames? You might want to try running
it without any specific configuration, so e.g. only
InitRef = NameService=corbaname::127.0.0.1 to see if it works.

What is your comandline for omniNames? (you should only specify -start
on the first invocation). You might also want to try to remove the
omninames log files and start omniNames again using -start once.

Hi,
WOW that was fast...
As far as the log file is concerned, i do see the /opt/log/omninames-<hostname>.[log|bak]
files. I also see an omniName.log file as I said.
Interesting that it should be non volatile by default, the one we are running is not! If I do a
namectl list before stopping, I see all the processes, after a stop start I get an empty list.
The config file is as follows, is there something that makes the service non volatile?
traceLevel = 0
traceExceptions = 0
traceInvocations = 0
traceInvocationReturns = 0
traceThreadId = 0
traceTime = 0
dumpConfiguration = 0
InitRef = NameService=corbaloc::tmxd12c:30000/NameService
DefaultInitRef = none
clientTransportRule = * unix,tcp,ssl
clientCallTimeOutPeriod = 90000
clientConnectTimeOutPeriod = 90000
thanks
Roger
-------- Original-Nachricht --------

Datum: Wed, 25 Jun 2008 14:53:40 +0200
Betreff: Re: [omniORB] Name service problems
omninames is non-valatile by default. It normally stores information in
/var/log/omninames-<hostname>.[log|bak] which needs to be writable by
the user running the process.
You can override this location by putting OMNINAMES_LOGDIR in the
environment or by specifying -logdir <directory> on the command line.
We're running omniNames out of daemontools (http://cr.yp.to) to make
sure it restarts on crashes - but we've yet to see one of these :)

Post by Roger wenham
Hi,
I am having problems with omniNames stopping for no apparent reason.
This happens about once a week, on various Solaris machines.
Has anyone else seen this behaviour, if so, I would be grateful for any

pointers

Post by Roger wenham
on how to go about debugging this.
The start command looks like this
/opt/omniORB/bin/omniNames -errlog /opt/log/omniName.log
There are no errors to be found in the omniName.log.
A second question is: When I re-start the name server after an error, it

has of course

Post by Roger wenham
forgotten all the names registered with it. I have searched the

documentation for the

Post by Roger wenham
way to make the name server non volatile, but have found nothing so far.
Can some kind soul enlighten me please...
Thanks in advance
Roger

[Duncan Grisby]

Post by Roger wenham
I am having problems with omniNames stopping for no apparent reason.
This happens about once a week, on various Solaris machines.
Has anyone else seen this behaviour, if so, I would be grateful for
any pointers on how to go about debugging this.

Run with command line arguments

-ORBtraceLevel 25 -ORBtraceInvocations 1 -ORBtraceThreadId 1 -ORBtraceTime 1

to get loads of debugging output. That might help track it down.

Post by Roger wenham
The start command looks like this
/opt/omniORB/bin/omniNames -errlog /opt/log/omniName.log
There are no errors to be found in the omniName.log.
A second question is: When I re-start the name server after an error,
it has of course forgotten all the names registered with it. I have
searched the documentation for the way to make the name server non
volatile, but have found nothing so far.

omniNames is _always_ persistent. There's no way to make it so it
isn't.

Perhaps you are starting it with a script that removes the
omninames-<hostname>.log and .bak files? That's where the persistent
information is stored.

Cheers,

Duncan.