topic Re: Tset problem in Operating System - Tru64 Unix

Tset problem

Vladimir Fabecic — Tue, 11 Oct 2005 06:20:24 GMT

Hello
I have very strange problem. On customer site I installed some RPM packages from freeware CD (lsof, zsh, ncftp, syd, top, monitor, vmubc, xfsm). I did it on two DS15 (single machines) and on Cluster (2 X ES47 + EVA3000). Also rebuilt kernels to include packetdriver. OS version is V5.1b PK5.
Everything is OK with both DS15, but on cluster machines there is a problem. When user logins, gets message: Type dtterm (or vt100) unknown.
It is still in test condition, so there is enough time.

Re: Tset problem

Vladimir Fabecic — Tue, 11 Oct 2005 06:24:45 GMT

Forgot to say: node which is booted second works extremly slow. First booted node works OK.

Re: Tset problem

Hein van den Heuvel — Tue, 11 Oct 2005 06:53:17 GMT

The message probably comes from the 'setup' program. (or possible 'resize').

Sounds like your termcap file ( /usr/share/lib/termcap ) is broken, or maybe there is something wrong with the softlink /etc/termcap whcih should point back to it?

The termcap file (man termcap) should have entries for vt220 and vt240 pointing on to vt200 for which the capabilities are defined on top of the vt100 stuff which is where the buck stops.

xterm has its own complete definition.

fwiw,
Hein.

Re: Tset problem

Hein van den Heuvel — Tue, 11 Oct 2005 06:56:04 GMT

I wrote "The message probably comes from the 'setup' program. (or possible 'resize')."

And then after the submit I noticed the title!
At first observation I had thought that 'tset' was a misspelling of test like I do myself. Now it is obvious it was simple a reference to the tool that reports the error. Sorry 'bout that.

Hein.

Re: Tset problem

Vladimir Fabecic — Tue, 11 Oct 2005 07:18:10 GMT

Very strange! Tset problem was fixed by coping termcap file from another node (DS15). Looks like file was corrupted! But how could it be corrupted?
And why is second node working so slow? Is there a bug?
Should I recreate the cluster?

Re: Tset problem

Ivan Ferreira — Tue, 11 Oct 2005 08:06:32 GMT

When we had problems with slow booting nodes in a cluster, the first booting fast and the others slow, was a memory channel interconnect, specially using lan interconnect.

Are you using Lan Interconnect or memory channel interconnect?

Check your lan interconnect, switch configuration, network adapter speed (set manually in the adapter and the switch), use lan_config, etc.

Re: Tset problem

Vladimir Fabecic — Tue, 11 Oct 2005 09:42:42 GMT

Hello Ivan
I also thought it was cluster interconnect. It is LAN (gigabit) interconnect (it is DEGXA-SB and fibre crossover cable, ant it worked OK till nowdays).
But "netstat -a -I bcm0" shows:
# netstat -a -I bcm0
Name Mtu Network Address Ipkts Ierrs Opkts Oerrs Coll
bcm0 1500 00:10:18:10:01:20 293079 0 293011 0 0
09:00:2b:02:01:04
09:00:2b:00:00:0f
01:00:5e:00:00:01
cf:00:00:00:00:00
bcm0 1500 DLI none 293079 0 293011 0 0
bcm0 1500 192.168.8 member2-icstcp0 293079 0 293011 0 0
all-systems.mcast.n
# netstat -a -I bcm0
Name Mtu Network Address Ipkts Ierrs Opkts Oerrs Coll
bcm0 1500 00:10:18:10:01:20 293795 0 293740 0 0
09:00:2b:02:01:04
09:00:2b:00:00:0f
01:00:5e:00:00:01
cf:00:00:00:00:00
bcm0 1500 DLI none 293795 0 293740 0 0
bcm0 1500 192.168.8 member2-icstcp0 293795 0 293740 0 0
all-systems.mcast.n

So there are no errors or am I wrong?
These card are set to 1000mbps fullduplex by console.
Did your nodes also work slow, not just booted slow?
I will change the crossover cable tomorow but I am not sure this would help.
Something cunfused me very much: how could termcap file get corrupted.

Re: Tset problem

Vladimir Fabecic — Wed, 12 Oct 2005 04:20:47 GMT

I think I have a big problem.
When I boot a node first (no matter what node), it boots fast and works fast. And when I boot other node, boot is extremely slow and works slow. When I shutdown first booted node, second starts to work fast. I changed cluster interconnect cable, but no change. Looks like cluster interconnect is working OK.
I was wrong when I said it is EVA3000. It is EVA4000 (sorry, sintax error). Both nodes have netrain interfaces for corporate network.
Any idea?

Re: Tset problem

Vladimir Fabecic — Wed, 12 Oct 2005 05:41:56 GMT

Looks like termcap is not the only corrupted file. Problems started after installing those RPM packages.
I do not know what happened. This should not have anything to cfs.
Does anyone know is there a bug or PK5 has some strange behavior?

Re: Tset problem

Vladimir Fabecic — Fri, 14 Oct 2005 07:54:18 GMT

Looks like I found the reason of problem. When I installed zsh on installation OS disk and executed zsh as root. Then had tset problem and file corruption. Never install zsh in cluster envirement!