[gentoo-user] ltsp woes

[gentoo-user] ltsp woes

[John Blinka]

Hi,

I help out with a local school using an ltsp server to run 50-60 ancient
pentium boxes as terminals.  I've run gentoo and ltsp on this server for 
almost
a year without problems.  However, a few weeks ago there had been a sudden
power outage (and no ups) which took the server and terminals down.
Ever since that event, we've not been able to get the terminals to boot.
We get as far as this message on the terminals:

Boot from (N)etwork or from (L)ocal? N
Probing...[3C5x9]3C5x9 board on ISA at 0x300 - 10baseT
Ethernet address: 00:A0:24:98:14:55
Searching for Server (DHCP)...
............ No Server found

The server is up and running.  I can log into it remotely via
ssh or VNC and everything seems to be operating normally.
A dhcp daemon is running, and, in fact is supplying ip addresses
to a bunch of windows boxes on the school network.

Any suggestions on how to debug this are welcome!

Thanks for your help.

John Blinka
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] ltsp woes

[Daniel da Veiga]

On 5/5/06, John Blinka <jblinka@neo.rr.com> wrote:
> Hi,
>
> I help out with a local school using an ltsp server to run 50-60 ancient
> pentium boxes as terminals.  I've run gentoo and ltsp on this server for
> almost
> a year without problems.  However, a few weeks ago there had been a sudden
> power outage (and no ups) which took the server and terminals down.
> Ever since that event, we've not been able to get the terminals to boot.
> We get as far as this message on the terminals:
>
> Boot from (N)etwork or from (L)ocal? N
> Probing...[3C5x9]3C5x9 board on ISA at 0x300 - 10baseT
> Ethernet address: 00:A0:24:98:14:55
> Searching for Server (DHCP)...
> ............ No Server found
>
> The server is up and running.  I can log into it remotely via
> ssh or VNC and everything seems to be operating normally.
> A dhcp daemon is running, and, in fact is supplying ip addresses
> to a bunch of windows boxes on the school network.
>
> Any suggestions on how to debug this are welcome!
>

How are the boxes connected to the network? A hub? A switch? Router
maybe? Try using the network spot (where one of the failing boxes is
connected) with a notebook or another machine that you know has no
problems. Also check the cables and other network devices between the
boxes and the DHCP server. I doubt a software failure would get all
your boxes down at the same time.

So, its my bet, a physical problem, but I'm not an expert.

--
Daniel da Veiga
Computer Operator - RS - Brazil
-----BEGIN GEEK CODE BLOCK-----
Version: 3.1
GCM/IT/P/O d-? s:- a? C++$ UBLA++ P+ L++ E--- W+++$ N o+ K- w O M- V-
PS PE Y PGP- t+ 5 X+++ R+* tv b+ DI+++ D+ G+ e h+ r+ y++
------END GEEK CODE BLOCK------

-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] ltsp woes

[John Blinka]

Daniel da Veiga wrote:

>
> How are the boxes connected to the network? A hub? A switch? Router
> maybe? Try using the network spot (where one of the failing boxes is
> connected) with a notebook or another machine that you know has no
> problems. Also check the cables and other network devices between the
> boxes and the DHCP server. I doubt a software failure would get all
> your boxes down at the same time.
>
> So, its my bet, a physical problem, but I'm not an expert.


A reasonable bet, but I'll have to apologize for omitting some information:
there are two rooms of terminals each connected via hub to
the network backbone.  Both rooms have the same problem, so it's probably
not network hardware.  I've seen this sort of problem before arise from
ltsp software configuration, so software problems at the server can 
bring all
the terminals down.  Between the time when the server was working and the
time it stopped working, I have probably upgraded some portion of the
ltsp package.  I'm wondering whether I inadvertently clobbered part of the
working configuration when I did so.

John Blinka
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] ltsp woes

[Uwe Thiem]

On 05 May 2006 19:57, John Blinka wrote:
> Daniel da Veiga wrote:
> > How are the boxes connected to the network? A hub? A switch? Router
> > maybe? Try using the network spot (where one of the failing boxes is
> > connected) with a notebook or another machine that you know has no
> > problems. Also check the cables and other network devices between the
> > boxes and the DHCP server. I doubt a software failure would get all
> > your boxes down at the same time.
> >
> > So, its my bet, a physical problem, but I'm not an expert.
>
> A reasonable bet, but I'll have to apologize for omitting some information:
> there are two rooms of terminals each connected via hub to
> the network backbone.  Both rooms have the same problem, so it's probably
> not network hardware.  I've seen this sort of problem before arise from
> ltsp software configuration, so software problems at the server can
> bring all
> the terminals down.  

Right. Somewhat besides your question: Are you really using hubs? I also seem 
to remember from your original post that the terminals are connected by 
10Mb/s which makes sense if you are using hubs. From my experience with 
server / thin client configurations, I would suggest to replace the hubs by 
switches and use 100Mb/s full duplex connections. That should boost the 
performance of your terminals by far. Anyway, this has nothing to do with 
your current problem.


> Between the time when the server was working and the 
> time it stopped working, I have probably upgraded some portion of the
> ltsp package.  I'm wondering whether I inadvertently clobbered part of the
> working configuration when I did so.

Please post your ifconfig output on the server and also the routing table.

Are you sure your server sees the dhcp requests at all? Run tcpdump on the 
appropriate interface on the server and boot just one terminal. Does the 
request come through?

Is dhcpd listening on the right interface? You mentioned a power outage. So 
some configuration could be messed up.

A power outage can also be accompanied by some power surges. That could fry 
the ethernet card in your server. Can you connect to it with a fat client? 
With dhcp? With a static IP configuration?

What happens if you plug a thin client directly into the server's ethernet 
card with a cross-over cable?

Uwe

-- 
Why do consumers keep buying products they will live to curse?
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] ltsp woes

[John Blinka]

Uwe Thiem wrote:

>Right. Somewhat besides your question: Are you really using hubs? I also seem 
>to remember from your original post that the terminals are connected by 
>10Mb/s which makes sense if you are using hubs. From my experience with 
>server / thin client configurations, I would suggest to replace the hubs by 
>switches and use 100Mb/s full duplex connections. That should boost the 
>performance of your terminals by far. Anyway, this has nothing to do with 
>your current problem.
>  
>

We originally used hubs and later moved to switches - my mistake.  But
it hasn't made much difference in performance.  Our performance bottleneck
is elsewhere.  But that's another discussion.

>>Please post your ifconfig output on the server and also the routing table.
>>    
>>
--> ifconfig
eth0      Link encap:Ethernet  HWaddr 00:02:A5:ED:2B:AD 
          inet addr:10.88.1.5  Bcast:10.88.1.255  Mask:255.255.0.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:382228 errors:0 dropped:0 overruns:0 frame:0
          TX packets:33183 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:40956294 (39.0 Mb)  TX bytes:9529879 (9.0 Mb)

--> route
Kernel IP routing table
Destination     Gateway         Genmask         Flags Metric Ref    Use 
Iface
10.88.0.0       *               255.255.0.0     U     0      0        0 eth0
default         ws510.ltsp      0.0.0.0         UG    0      0        0 eth0

>>Are you sure your server sees the dhcp requests at all?
>>
Yes.  The system logs are full of requests.  The ltsp server is also a 
dhcp server
for the entire campus, and all of the non-ltsp-terminal boxes are 
functioning
properly on the network.

>>Run tcpdump on the 
>>appropriate interface on the server and boot just one terminal. Does the 
>>request come through?
>>    
>>
Can't do that right now - don't have physical access to the machines on 
the weekend
and can't boot a terminal now.

>>Is dhcpd listening on the right interface? You mentioned a power outage. So 
>>some configuration could be messed up.
>>    
>>
There's only one interface, eth0.  Here's the command that started dhcpd 
- looks like it
should be listening to eth0.

--> ps aux | grep dhcp
dhcp      9009  0.0  0.0   2680  1656 ?        Ss   May04   0:01 
/usr/sbin/dhcpd -q -pf /var/run/dhcp/dhcpd.pid -user dhcp -group dhcp -q 
eth0

>>A power outage can also be accompanied by some power surges. That could fry 
>>the ethernet card in your server. Can you connect to it with a fat client? 
>>With dhcp? With a static IP configuration?
>>    
>>
It's not fried.  I can ssh from a remote location into the server, and 
the server has only one
ethernet card in it.

So, the hardware is working, and dhcpd is filling the system logs with 
stuff like

May  6 14:55:29 [dhcpd] DHCPDISCOVER from 00:a0:24:98:14:55 via eth0
May  6 14:55:30 [dhcpd] DHCPOFFER on 10.88.3.122 to 00:a0:24:98:14:55 
via eth0

which means that dhcpd is running, listening on eth0 and talking to ltsp 
terminals.
But, judging from the ltsp-terminal error messages I posted originally, 
the terminals
don't seem to think that they've made a satisfactory contact with dhcpd 
on the server.


John Blinka
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] ltsp woes

[Uwe Thiem]

On 06 May 2006 20:26, John Blinka wrote:
> Uwe Thiem wrote:

> We originally used hubs and later moved to switches - my mistake.  But
> it hasn't made much difference in performance.  Our performance bottleneck
> is elsewhere.  But that's another discussion.

Alright, but I would really like to discuss this further. Maybe off-list when 
your current problem is solved.

>
> >>Please post your ifconfig output on the server and also the routing
> >> table.
>
> --> ifconfig
> eth0      Link encap:Ethernet  HWaddr 00:02:A5:ED:2B:AD
>           inet addr:10.88.1.5  Bcast:10.88.1.255  Mask:255.255.0.0
>           UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
>           RX packets:382228 errors:0 dropped:0 overruns:0 frame:0
>           TX packets:33183 errors:0 dropped:0 overruns:0 carrier:0
>           collisions:0 txqueuelen:1000
>           RX bytes:40956294 (39.0 Mb)  TX bytes:9529879 (9.0 Mb)

Your broadcast address is wrong. With that netmask, it must be 10.88.255.255.

>
> --> route
> Kernel IP routing table
> Destination     Gateway         Genmask         Flags Metric Ref    Use
> Iface
> 10.88.0.0       *               255.255.0.0     U     0      0        0
> eth0 default         ws510.ltsp      0.0.0.0         UG    0      0       
> 0 eth0

If you have only one interface, what is that default route good for? Not that 
I can see any harm from it.

Try to correct the broadcast above. Maybe that will solve the problem already, 
although it's hard to understand how you fat clients can work with it.

If the broadcast doesn't solve it, please post your /etc/dhcp/dhcpd.conf as 
well.

Uwe

-- 
Why do consumers keep buying products they will live to curse?
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] ltsp woes

[John Blinka]

Uwe Thiem wrote:

>Try to correct the broadcast above. Maybe that will solve the problem already, 
>although it's hard to understand how you fat clients can work with it.
>  
>
I corrected the broadcast address, but it didn't help.

>If the broadcast doesn't solve it, please post your /etc/dhcp/dhcpd.conf as 
>well.
>  
>
Here it is:

default-lease-time           421600;
max-lease-time               421600;
ddns-update-style none;
allow booting;
allow bootp;
authoritative;
#boot-unknown-clients off;

option option-128 code 128 = string;
option option-129 code 129 = text;

option subnet-mask            255.255.0.0;
option broadcast-address      10.88.255.255;
option routers                10.88.1.254;
option domain-name-servers    10.88.1.3;
option domain-name            "elms.k12.oh.us";
option root-path              "10.88.1.5:/opt/ltsp-4.1/i386";
option netbios-name-servers   10.88.1.3;
option netbios-node-type      8;

shared-network WORKSTATIONS {
  subnet 10.88.0.0 netmask 255.255.0.0 {
     range dynamic-bootp 10.88.3.1 10.88.4.254;
     use-host-decl-names       on;
     option option-128     e4:45:74:68:00:00;
     option option-129     "NIC=3c509";

     option log-servers        10.88.1.5;

     # trick from Peter Rundle <peter.rundle@au.interpath.net>
     if substring (option vendor-class-identifier, 0, 9) = "PXEClient"
     {
        filename      "/lts/pxe/pxelinux.bin";
          # NOTE: kernels are specified in /tftpboot/lts/pxe/pxelinux.cfg/
     }
     else
     {
        filename        "/lts/vmlinuz-2.4.26-ltsp-2";
     }
  }
}


-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] ltsp woes

[Uwe Thiem]

On 08 May 2006 19:12, John Blinka wrote:
> Uwe Thiem wrote:
> >Try to correct the broadcast above. Maybe that will solve the problem
> > already, although it's hard to understand how you fat clients can work
> > with it.
>
> I corrected the broadcast address, but it didn't help.
>
> >If the broadcast doesn't solve it, please post your /etc/dhcp/dhcpd.conf
> > as well.
>
> Here it is:

Can't see where it breaks. :-(

Uwe

-- 
Why do consumers keep buying products they will live to curse?
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] ltsp woes - problem solved

[John Blinka]

The /etc/dhcp/dhcpd.conf file needed another line:

next-server  ip-address-of-server;

with the upgrade to dhcp-3.0.3 (and maybe a version or
two earlier).  With that addition, all is now well.

John Blinka
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] ltsp woes - problem solved

[Uwe Thiem]

On 09 May 2006 19:15, John Blinka wrote:
> The /etc/dhcp/dhcpd.conf file needed another line:
>
> next-server  ip-address-of-server;
>
> with the upgrade to dhcp-3.0.3 (and maybe a version or
> two earlier).  With that addition, all is now well.

Thanks for hitting this before me! ;-)

Uwe

-- 
Mark Twain: I rather decline two drinks than a German adjective.
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] ltsp woes

[Uwe Thiem]

On 06 May 2006 20:26, John Blinka wrote:
> Uwe Thiem wrote:
> >Right. Somewhat besides your question: Are you really using hubs? I also
> > seem to remember from your original post that the terminals are connected
> > by 10Mb/s which makes sense if you are using hubs. From my experience
> > with server / thin client configurations, I would suggest to replace the
> > hubs by switches and use 100Mb/s full duplex connections. That should
> > boost the performance of your terminals by far. Anyway, this has nothing
> > to do with your current problem.
>
> We originally used hubs and later moved to switches - my mistake.  But
> it hasn't made much difference in performance.  Our performance bottleneck
> is elsewhere.  But that's another discussion.

Hi John, I would like to come back to you on this off the list. Unfortunately, 
all mails to your email address bounce with "server unreachable". Do you have 
another email address?

Uwe

-- 
Mark Twain: I rather decline two drinks than a German adjective.
-- 
gentoo-user@gentoo.org mailing list