swinstall problem

[Stephen Keane]

I had been using swinstall for a couple of days OK. Then today ...

* Begining Selection
* Target connection succeeded for "foo:/".
ERROR: RPC exception: "Connection closed (dce/rpc)" 07/04/05 14:29:53 BST
ERROR: A Remote Procedure Call to an agent has failed. Could not start a management session for
"foo:/depot_dir/foo.depot". Make sure the host is accessible from the network.
* Source connection failed for
"foo:/depot_dir/foo.depot".
WARNING: More information may be found in the daemon logfile on this
target (default location is foo:/var/adm/sw/swagentd.log).
* Selection had errors

I've changed the host name /depot path top protect the innocent. I've checked the host name / IP using nslookup. I've also checked /etc/hosts and /etc/nsswitch.conf

Attached is swagentd.log

Any clues as to where to start looking?

[Luk Vandenbussche]

Try to restart swagentd

/sbin/init.d/swagentd stop
/sbin/init.d/swagentd start

[Mark Nieuwboer]

Did you start the swagent daemon with swagentd -r on the server you want to install.
It looks like the daemon is down.

grtz. Mark

[Stephen Keane]

This is all taking place on the same machine. The software depot is in a local directory. I have rebooted the machine and started/stopped swagentd (as suggested).

It looks like swagentd is coring (SIGNAL 11), but I don't know why. Is there an RPC log somewhere I could consult?

[Mark Nieuwboer]

what does the log /var/adm/sw/swagentd.log say.

grtz. Mark

[Stephen Keane]

rp3440 running HP-UX 11.11

[Mark Nieuwboer]

Stephan,

Find this from an old forum.
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=112523

[Stephen Keane]

Thanks for that thread. I don't have either glance or measureware running. I will try the rpc_binding flag, though I don't know why I should have to!

[Stephen Keane]

The rpc_binding_info flag made no difference. swagentd is still core dumping SIGSEGV.

[Biswajit Tripathy]

See if you have any log in /var/opt/dce/svc files.

- Biswajit

[Ermin Borovac]

Have a look at KBRC00016523. It may be relevant if you have the latest SD-UX patch (PHCO_32116).

[Robert-Jan Goossens_1]

Stephen,

Check this doc.

http://www4.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000066922064

Document description: SYS ADM: Swinstall error- Attempt to create the job failed.
Document id: A5029755

Failure registering as remote procedure call server using
protocol sequence ncacn_ip_tcp with endpoint 2121. No
addresses dce/rpc.

To solve this problem, execute the following commands:

1. cd /var/adm/sw/queue

2. mkdir /tmp/queue

3. find . -print | cpio -pdmux /tmp/queue

4. rm -rf /var/adm/sw/queue/*

HTH,
Robert-Jan

[Stephen Keane]

Biswajit,

I checked the directory, there are log files, but they are all empty (zero bytes).

[Stephen Keane]

Ermin,

I don't know what KBRC00016523 is?

[Stephen Keane]

Robert,

I tried the fix you suggested, but the problem remains. Should I delete /tmp/queue now?

[Robert-Jan Goossens_1]

Hi Stephen,

I don't think the /var/adm/queue dir holds vital info.

Is the /depot_dir directory a nfs filesystem ? (firewall ?)

Regards,
Robert-Jan

[Stephen Keane]

The depot I'm trying to install is local, no NFS/Samba or anything like that. I did think about patches, and ran a patch assessment. I downloaded the latest Gold bundle (June) some 450 odd MBytes, thanks HP!

Curiously, I had no problem installing the patch bundle, though that was a depot too.

Maybe the depot I'm using has a problem? Is there any way of checking it for integrity?

[Greg Vaidman]

to verify the integrity of the depot, run swverify -v "*" @ /path/to/depot

[Mateja Bezjak]

See if the depot is registered?
swreg -l depot

Regards,
Mateja

[Stephen Keane]

Greg,

I ran the swverify command you suggested, but it failed sying there was no installed software at the depot location. Which there isn't, as it won't bloody install it!

[Stephen Keane]

Mateja,

swreg seems happy enough, no errors.

[Greg Vaidman]

can you get a swlist of the depot? ("swlist -d @ /path/to/depot", and post the output for us)

also, from your original message, it looks like you're using a file (as opposed to a directory) depot. since a file depot is just a tar file, can you also post the output of "tar tvf foo.depot"?

[Ermin Borovac]

Stephen, sorry KBRC00016523 is document ID in the HP's technical knowledge base article. If you have support contract with HP go to www.itrc.hp.com, search by 'doc ID', enter KBRC00016523 and you'll find issue very similar to yours. It's only relevant if you have PHCO_32116.

[vinod_25]

hi stephen

The swinstall "rpc exception communication error" most likely indicates a hostname resolution problem.

check the hostname and ip address in
/etc/resolv.conf, /etc/nsswitch.conf or
/etc/hosts

The /usr/sbin/swagentd daemon, which is required when using SD-UX commands like swinstall, keeps hostname resolution information from the time it is initialized. Network or hostname resolution changes can cause the static information used by the swagentd to become outdated and incorrect
and can cause the "rpc exception communication error" exhibited in this case.

issue /usr/sbin/swagentd -r to restart the swagent daemon...

even if that doesnt solve the problem try
- lanscan
- ifconfig (for each network interface card)
- netstat -nr
- nslookup `hostname`
- nslookup IP_ADDR (where IP_ADDR is IP returned by nslookup `hostname`)

you can find additional info from /var/adm/sw.

Regards

Vinod

[Stephen Keane]

vinod,

all the network stuff looks fine. host-ip and ip-host resolution seems OK. The thing is I can install patches, stored locally on the machine, but not a particular depot. Which leads me to believe the depot is faulty.

Greg,

the swlist command produces ...

# Initializing...
# Contacting target "foo"...
Memory fault(coredump)

file core

core: core file from "swlist" - received SIGSEGV