Inetd dying

[Alexander M. Ermes]

Good day, everybody.
During the last days i had two ( as i think unmotivated ) stops of the inetd.
I work on this problem with HP support, but i would like some independent opinion on that.
We have installed the necessary patches on our system.
V2500
HP-UX 11.00
PHSS_24303 installed and patchlist attached.
We tried to install patch PHSS_24301, but on the analysis run the filesets were skipped.
Here the log :

======= 03/14/02 14:42:30 MET BEGIN install AGENT SESSION (pid=17260)

* Agent session started for user "root@hpnss-01". (pid=17260)

* Beginning Analysis Phase.
* Source: hpnss-01:/var/tmp/patches/PHSS_24301.depot
* Target: hpnss-01:/
* Target logfile: hpnss-01:/var/adm/sw/swagent.log
* Reading source for product information.
* Reading source for file information.
NOTE: The patch fileset "PHSS_24301.LANG-HELP,r=1.0" may only be
installed upon a previously installed base fileset. The
specification for this base fileset is
"Auxiliary-Opt.LANG-HELP,fr=B.11.00,v=HP". Since there is no
fileset on the target or selected from the source which
matches this specification, fileset
"PHSS_24301.LANG-HELP,r=1.0" will be excluded.
NOTE: The patch fileset "PHSS_24301.LANG-STARTUP,r=1.0" may only be
installed upon a previously installed base fileset. The
specification for this base fileset is
"Auxiliary-Opt.LANG-STARTUP,fr=B.11.00,v=HP". Since there is
no fileset on the target or selected from the source which
matches this specification, fileset
"PHSS_24301.LANG-STARTUP,r=1.0" will be excluded.
* Executing preDSA command.

* Summary of Analysis Phase:
* 2 of 2 filesets will be Skipped.
* The Analysis Phase succeeded.


======= 03/14/02 15:16:59 MET END install AGENT SESSION (Canceled)
(pid=17260)

* 03/14/02 15:22:02 MET BEGIN list AGENT SESSION (pid=25375)
* Agent session started for user "root@hpnss-01". (pid=25375)
* 03/14/02 15:39:14 MET END list AGENT SESSION (pid=25375)

The previously installed fileset is part of PHSS_24303 and this one is installed.
Did anybody run into a similar problem and how did you solve it ?
I really appreciate help in any case.
Rgds
Alexander M. Ermes

[Roger Baptiste]

hi,

What was the last message recorded in the syslog file and dmesg output with relation to inetd dying??

I remember having this problem couple of times in a V2500 box on 11.00, but that was more than two years back. But it happened twice and afterwards never reoccured.

If the problem is occuring frequently, you could probably run trace on it. But that would require space. I also think, there is a version of inetd which has a debug flag to record more messages.

Sorry, dont have any more info.

-raj

[Alexander M. Ermes]

Hi there.
Here the records from syslog.log :

Mar 14 00:30:27 hpnss-01 ftpd[558]: FTP session closed
Mar 14 00:33:50 hpnss-01 ftpd[6736]: FTP LOGIN FROM 153.2.18.217 [153.2.18.217], ftpnew
Mar 14 00:33:53 hpnss-01 ftpd[6736]: FTP session closed
Mar 14 02:42:38 hpnss-01 inetd[890]: Going down on signal 11


Mar 14 15:51:23 hpnss-01 ftpd[24083]: FTP session closed
Mar 14 15:51:23 hpnss-01 inetd[6492]: Going down on signal 11


Rgds
Alexander M. Ermes

[Ron Kinner]

http://bizforums.itrc.hp.com/cm/QuestionAnswer/0,,0xa4bb28e43106d6118ff40090279cd0f9,00.html

Refers to the same problem. Sig 11 is something about a memory segment violation.

Ron

[Helen French]

Hi Alex,

Did you install all dependency patches and the base file set ?

The "inetd dying on signal 11" is a memory related error.

Check this thread, this refers to the same problem:

http://us-support.external.hp.com/cki/bin/doc.pl/sid=3813dc7609d7f31491/screen=ckiDisplayDocument?docId=200000059171595

HTH,
Shiju

[S.K. Chan]

Signal 11 is a memory segment violation. HP response center is best to handle this. They have a version of inetd that can do debugging, ask for that and they should be able to install and eventually find out the root cause. It's going to be tough to pinpoint the root cause though if inetd only dies occasionally, meaning this is not a frequent re-occuring problem.

[Roger Baptiste]

$ grep 11 /usr/include/sys/signal.h
# define SIGSEGV 11 /* Segmentation violation */

Shows something to do with inetd running into segmentation error. Not normal. I think HP deserves a call ;-)

cheers
-R

[A. Clay Stephenson]

Hi Alexander:

While this is not meant to be a fix (and you should find the problem), because inetd is such an important daemon, I always install a cronjob that runs every 5 minutes to check the status of inetd and restart it if necessary.

You might consider making a crontab entry that runs this script, 'inetdchek.sh'.

Regards, Clay

[Ralf Hildebrandt]

I recommend not using inetd at all:
It has no provisions agains resource exhaustion,
puts no limits on concurrent connections
and, as you noticed, crashes.

Try xinetd or tcpserver instead.

[harry d brown jr]

Alexander,

I know this sounds almost stupid, but check your memory kernel parameters and your swap space. And I'd try using the custom patch manager to make sure you have any dependency patches.

live free or die
harry

[Joseph Loo]

Alexander,

Do you have HP C/Ansi C Developer's Bundle (B3901BA) install in your server or B3989BA for your workstation? I believe you would need this before you can carry out patch PHSS_24301.

However, the above product would need a codeword before you can install it?

Hope this help.

Regards,
Joseph.

[Alexander M. Ermes]

Hi everybody.
Thanks for help.

I have contacted HP, installed the debugging
version of inetd and wait for a core file to be created. Then i will send this to HP for analysis.
Clay, your watchdog script is a real goodie.
HP provided this one :

#!/sbin/sh
cd /
while sleep 60; do
if /usr/sbin/inetd >/dev/null 2>&1 && [[ -f /core ]]; then
mv /core /var/tmp/core-inetd.$(date "+%y%m%d-%H%M%S")
echo "inetd dump in /var/tmp/!" | mailx -s "inetd dump!" root
fi
done

Joseph,
i do not have this on my server, because we usually do not use it.

Ladies and gentlemen, thanks for the help given until now.

Kind regards
Alexander M. Ermes

[Joseph Loo]

Alexander,

I believe HP will reply you by asking you to install that product as I have exactly the same swagent log message when I try to install that patch, PHSS_24301.

Regards,
Joseph.

[Alexander M. Ermes]

Hi Joseph.
If HP does so, i will them, that i got along without it for the last 4 years and why it should not stay that way.
If they will try to enforce something, there are other vendors for server systems.
Rgds
Alexander M. Ermes

[Alexander M. Ermes]

Thanks for your efforts.