NFS START HALTS SYS BOOT

[uspfoms]

I have a D380 sys running HP 11.0 it's hanging when booting at the START NFS CORE SUBSYSTEM message.
How can I turn NFS off ? When I can't get into the sys. I have a ignite tape for this box, but can't find where you can turn off NFS in
ignite setup.


Thanks

[Patrick Wallek · Accepted solution]

You can't turn it off from within Ignite.

What you can do is boot the system, interrupt the boot at the "press any key in 10 seconds" prompt. Then type 'boot pri' and say 'Y' to "Interact with IPL". Then type 'hpux -is' to boot into single-user mode.

Once there, type 'mountall' to mount the vg00 filesystems (if you have other filesystems and you get errors, don't worry). Now you need to edit the /etc/rc.config.d/nfsconf file to turn off all NFS.

Once you edit that file, reboot and now NFS should not start anymore.

[Tim Adamson_1]

Hi,

You can do what Patrick suggested or try pressing CTRL \ to halt the startup in its tracks. You will be presented with a login prompt and should be able to login (it will be jumbled in with whatever is displayed on the screen and you may need to press return several times to actually see it). However, the machine will not have completed its bootup. It will have stopped at NFS. Edit the files in /etc/rc.config.d to disable NFS from starting and reboot again.


Cheers!

[Pete Randall]

Since this seems to be the more popular of the two identically title threads, I'll copy my comments here - then maybe someone of HP persuaion can get rid of the other post.




99 times out of 101, this is the result of a host name resolution issue. To get by it in the meantime, either boot into single user mode, or interrupt the boot process when it gets to the NFS start by pressing the break keys (if I recall correctly). You then need to straighten out you /etc/rc.config.d/nfs.conf file. You can edit it to not start NFS in the meantime, but, ultimately, you are probably going to have to resolve your hostname resolution issues.


Pete

[uspfoms]

Thanks guys
I think this will fix my problem,
if I can get vi to work properly
it's acting strange in single user mode ?



Thanks Again

[Pete Randall]

Set your TERM variable to hp or vt100 (preferably hp):

export TERM=hp

That should straighten out vi.


Pete

[Bill Hassell]

Just s note about vi:
problems with vi relate to the terminal you are using and the TERM variable. In the good old days, every HP-UX server came with a 'real' HP console and as you booted up, even in single user mode, the terminal would be identified automatically using ttytype. However, today's sysadmins are using a LOT of imitations, non-HP dumb terminals or worse, cheap terminal emulators that run on PCs. Most of the dumb terminals and PC emulators attempt to emulate the prehistoric DEC VT100 terminal, now seen only in museums and some government offices :-)

To complicate a fairly simple task (identifying the terminal), the entire rp-series, SuperDomes and older A-, L-, and N-class servers have a multi-access GSP subsystem in which the designers tried to handle multi-port access (remote support modem, console, built-in LAN console) and decided to store a hard-coded ID inside the GSP. This ID has nothing to do with the 'real' terminal(s) attached to the console ports except by accident. If the GSP is set to hpterm, then the ttytype (and tset) commands will not poll the terminal for an ID but simply assume the ID from the GSP, and thus, TERM=hp will be set *PRIOR* to running /etc/profile. Well and good as long as you have a 'real' HP terminal or emulator connected to the console.

However, if you have the GSP set to vt100 and are using an HP terminal, vi (and SAM and swinstall, etc) will all go crazy since TERM does not match the real terminal. The opposite is also true: hpterm in the GSP versus a dumb vt100 as the console...vi will fail.

The fix is to pick a standard console terminal and then make the GSP match using the CA command. I recommend using an HP terminal or emulator and setting GSP to hpterm. The difference between vt100 and hpterm when running SAM or swinstall is phenomenal. I also recommend deleting the code /etc/profile that bypasses ttytype when TERM is already set. TERM gets set by telnet based on TERM in the client machine, and this value may not be useful, so always query the terminal.

Unfortunately, ttytype and tset bypass the query code if the connection is through the GSP console path.

[uspfoms]

OK, guys finally got it past the NFS halt,
Now it locks up at the start of the CDE gui.
?????????


Thanks

[Patrick Wallek]

OK, this definitely now sounds like a hostname lookup issue.

Are you using DNS to lookup hosts? Is this host defined in its local /etc/hosts file correctly? Do you have an /etc/nsswitch.conf file and a /etc/resolv.conf file if you are using DNS?

The contents of the /etc/nsswitch.conf file, at least the hosts line, should look like:

hosts: files [NOTFOUND=continue UNAVAIL=continue] dns

I always like to put files first and then DNS.

[Pete Randall]

As I said originally: hostname resolution!


Pete

[uspfoms]

Yes Pete it is a HOST NAME problem,Here's the whole story.
This started after our network folks upgraded our switch (1GB). Since this is a TEST box I did'nt know it was having problems until a few days later. This sys only has a 10Mb card Since I only use this to test with. Everything on the Cisco switch looks ok. I've tried setting the switch to 10/full and 10/half no luck with either. I've checked/created all the previously mentioned files. Still hanging at the GUI startup.
Anymore ideas ?


Thanks

[Pete Randall]

Have you tried nslookup? By hostname? By IP address? Can you ping this box from another machine by both hostname and IP? Check uname output against /etc/hosts and against your nslookup results.


Pete

[uspfoms]

Pete, nslookup by ip/hostname returns correct info. I can not ping this box from other machines. uname -a returns "UNKNOWN" in host name field. ?????????????????????????



Thanks again !!!!!!

[Pete Randall]

I would suggest running "/sbin/set_parms hostname" to (re)set the hostname. It will prompt you to re-boot and you should. Afterwards, I think things should be correct. If not, we'll pursue it from there.


Pete

[Patrick Wallek]

What IS the hostname of this box? Is it more than 8 characters long? If so, change it so that it is 8 characters or less. The 'uname -a' returning UNKNOWN is probably the majority of the problem. You might also check your /etc/rc.config.d/netconf file and make sure the hostname is set there.

Also make sure that there are NO backup copies of the netconf file in /etc/rc.config.d if there, delete them or move them somewhere else.

[uspfoms]

Thanks pete, The set_parms addl_netwrk command
showed the wrong DNS setup. Got that changed and finally now I can ping the system.

Problem is it still is hanging up at the
(CDE) common desktop environment .....
any ideas ????

Thanks again and again for all the help !
This forum is the best thing HP has going !

[Bill Hassell]

I would start by renaming /etc/resolv.conf and /etc/nsswitch.conf temporarily and make sure that /etc/hosts has the correct hostname and IP address. ping occurs at a very low level while CDE requires a lot of (correctly configured) high level networking to function. Use nslookup for both the name and also for the IP address. Checking both (and finding them equal) should correct the problem.

Once simple /etc/hosts works, then add /etc/resolv.conf (while running) and again check your own IP and hostname. This time it will only use the nameserver. If it fails, then find out why your DNS box is not reachable. The last step is to put /etc/nsswitch.conf back in place and perhaps adjust the rules (see man switch and also man nsquery).