Bizarre Post-boot Behavior

[Eldee Stephens]

Folks,

We just did a cold install of HP-UX 11i v1.6 on our rx2600 here, and are encountering absolutely bizarre system behavior. The system cannot establish a network link, so to avoid timeouts, we disabled the NFS client and server, as well as the SNMP daemons.

Now after a very long boot, it eventually just presents a console login prompt. After which we logged in (surprised it never fired up CDE), and a few minutes later were presented with:

Starting Desktop Login on display x...
Wait for the Desktop Login screen before logging in...

So we waited and waited, and nothing happened. Checking active processes, I didn't even see xinit. Then, while trying to look at some config files, the screen went completely black. No amount of mouse clicking or keyboard activity accomplished anything. Then four minutes later, the console returns exactly where it was before. We worked for another few minutes, and then the screen went black again.

What the hell is going on here? Is X11 trying to startup but can't? What can I do to move forward? I really, really need this damn thing up very soon. This behavior was occuring before the cold install, which was what prompted us to do it. Some weeks ago, everything was fine: the system booted normally into the CDE login screen, network activity was fine. Then we tried using SAM on day, the system froze, and ever since we've been unable to get CDE running or the network config to work correctly.

Has anyone ever seen anything like this?

-- eldee

[Steven E. Protter]

A charactarstic of HP-UX if there are servere problems is that very long boot.

The last time it happened to me, there was an error in network configuration. I goofed up and put an extra dot in /etc/rc.config.d/netconf

Since you did a cold install, the problem is more likely a problem with network mask or some other portion of networking.

Also note that HP-UX does not like two NIC cards on the same subnet. That will lock up networking tight.

So, check the network configuration, perhaps comment out all but one card and try again.

Though your startup probably hung on the NFS step, I'm thinking the problem is networking.

SEP

[Eldee Stephens]

Steven,

You're absolutely correct in assuming networking issues. I went ahead and disabled the NFS client/server, the SNMP daemons, and the software install daemon because if they can't poll the network, they hang and never truly timeout. So to move further in the boot process, that's what I did.

I figured that SAM would be a safer way of fixing the networking issues than editing all of the conf files in rc.config.d -- but if you can't even get X11 started....

What I'm wondering is if there is something else that could be at issue here. Surely a bad IP address would not cause this behavior -- or would it?

-- eldee

[Eldee Stephens]

Folks,

Just had a thought... could the unusual system behavior -- X11 unable to start -- be related to the CDE default mount point being /nfs ? If I recall correctly, there is some variable somewhere like DTMOUNTPOINT that defaults to /nfs, and if there is nothing in /nfs since the client and server are disabled, Xsession would never be able to start...

-- eldee

[Patrick Wallek]

This is very likely network related.

What happens if you try to do an:

# nslookup server_name

# nslookup x.x.x.x
where x.x.x.x is the servers IP address

Are you using DNS or is everything coming from the /etc/hosts file?

If you are using DNS, did you set up an /etc/nsswitch.conf file and /etc/resolv.conf file.

Check your /etc/nsswitch.conf file and make sure the hosts line is something like:

hosts: files dns

[Brian Sisk]

I've seen this behavior on systems that are trying to display X where only a tty console device exists. If that's the problem you can edit the /etc/dt/config/Xservers file and comment the 'local@console[...]' line (usually the last line in that file) and then execute /var/dt/bin/dtconfig -reset.

[Steven E. Protter]

SAM is not safer for networking. Its happened to me.

It sometimes puts double entries into /etc/hosts which can directly caus your current behavior.

All the files you need to look at are already listed in this thread.

An idea light just lit up over my head.

See it?

This was happening before the cold install.

Bing!

What changed when this started.
DNS change.
Switch change
change to /etc/nsswitch.conf
Did someone swap out a switch?
Are your switch ports explicit or autonegotiate? HP NIC's like explicit settings.

I suppose this could even be the network card.
We had a bad hub that was freaking out a built in LAN card on our rp5450 server. All we did was replace the hub and a bunch of really funky networking problems went away.


SEP

[Eldee Stephens]

I went ahead and did another cold install, but this time did not enable any of the ethernet cards. I'll bring up the interfaces later today and begin diagnosis, using your suggestions. We are using DNS, but that wasn't really the problem. We couldn't even ping a neighbor on the same subnet.

So I'll rigorously go the process of testing the cards, then I'll test loopback, etc., etc. Hopefully I'll nail it down soon enough. And I'll also try editing the dt config files before I try starting Xsession if I'm launching from the console. Thanks for everyones' help....

-- eldee

[Angus Crome]

A couple of things to add;

1) Check your DNS (reverse lookup should return the same name).
2) Check the network cable.
3) You can alias your machine name on the loopback line in /etc/hosts and restart networking. You should at least be able to get CDE up and working that way. Make troubleshooting much easier.
4) Make sure /etc/nsswitch.conf goes to files first. This allows your machine to know itself, even when networking is (not working).
5) Check the ports for equivalent setting between your NIC and the Switch (no autonegotiation, unless this is Gigabit Fiber), same speed and duplex.

This is definitely a networking problem.

[Sorrel G. Jakins]

On our rx2600 we put in a graphics console as well as the tty.

Just a thought...