Crack rack is operational...

[Michael]

Russ - where did the rc5 load itself? Is it running off of the floppy or did you copy it onto the ram drive? Since you should be running a root on each of the crack racks, I'm scratching my head as to why you have an access error. I don't have rc5 running on my linux machine (it's doing SETI), so I'm not 100% certain how the Linux version handles the files. Linux is picky about who can get access to what files.

Michael

I would want to run everything on the ramdisk. You would install the rc5 client onto the RAM disk and run it from there. It looks like the rc5.out program is in the root directory and the program is not associated with root for some reason.

Sorry - I need to go to bed. Try posting in the OS form and steer linux people to this thread.

 

[ccalvert]

Still nothing from dnet, Russ. Excuse me while I lose my mind waiting.

I think that if I don't here from them extremely soon, I am going to have to send another e-mail. I would just send it right now, but if I am going to attatch a floppy image to it, I would like to work on my scripting a little bit, so as to make it somewhat less laughable

In the meen time, I keep waiting... and waiting... and waiting...

 

[Russ]

Thanks, Michael.

I got the writing error solved; it's saving the completed packets in buff-out now. Still getting network open error, though. Think I'll head home and worry about it tommorow.

Russ, NCNE

 

[ganerc]

Hi Russ,

ur error: Network::Failed to resolve name "euro.v27.distributed...

has something to do with ur dns.
Put in /etc/resolv.conf : nameserver
and then the ip-adress of ur dns-server and everything should work fine.
But after this u have to reboot ur box(es).

cy@

 

[cory]

question?

on your boot floppy do you just have a rom image
or do you have a linux kernel?

it is much easier to just have a bootrom and then
tftp the kernel down to the machine and
then have it mount its root filesystem with nfs.

that way each node can use the exact same boot floppy
and it is never written to.

all config and filesystems are on the main linux box.

and i have a script that when i want to setup up a node
i just run the script and it sets up a new root filesystem
and then i just edit about a couple of files and
turn the node on.

from then on you can make changes to every node
from the main one.

one of the things that i did with my seti stuff was to write a
script that would look at the state.sah file for each of the nodes
and display info on how long it had been running,
# of hours to complete w/u, when the last time it
updated the state.sah file and some other stuff.
that way i can take one quick look and see if any node is
hung for whatever reason (like yesterdays problem at
the main setiathome site)
next step is to hook up my x10 stuff so that if the node
is hung that the main linux machine will cycle the power on that
node.

 

[cory]

if you want a boot rom for a specific nic card i got a bunch cause
when i built mine it built one for every card.

building the kernel image is not that tuff either.

setting up dhcp to do the tftpboot command is like one extra line
and the nfs share stuff is very simple as well.

just let me know if you want any specific info.

(i will have a total of 20 processors running at home
with a total of about 15000 mhz, 3 running windows all the rest
running linux)

 

[Michael]

Cory - here is the hardware info for Russ' crack rack:

http://www.compucheap.com/pages/crackrack.html

He is very new to Linux. I bet he would love a floppy that could boot into Linux and then allow him to use NFS to and from his server from each of his nodes. With telnet from his server, he really shouldn't need a monitor and keyboard for each node (he's using a switch box right now).

I don't know enough about the Linux Router disk that he's using to help much more than general linux advice. From reading your SETI thread, I think you have a linux cluster running and Russ has asked for well explained help (because he is just getting his feet wet with Linux) in previous threads.

Russ - You can change the config file for RC5 to go to a keyserver that can be resolved. Are you running a pproxy on one of your network machines or do you connect to Mika's? Is your router and then your server passing the ports that dnet communicates on to the nodes? The can't resolve error sounds like a DNS problem, but it might be a porrly worded error message from the program.

Michael

ps - I knew you could help, that's why I asked you to come visit here!

 

[cory]

i need more info on that nic he is using.
is it an ne2000 clone?.

i have bootroms that can be put on floppy for about
50 different nic chipsets.

i wish i had an eprom burner cause the cards i use have a bootrom
socket then i could get rid of the floppy completely.

i use the dlink 530tx+ which is a 10/100 pci card
that you can get for $13.

oh and btw i dont use video cards or keyboards.
(i just use the video card to do the bios setup
then test it and then pull the video card)

 

[Batti]

cory - do keep posting information. I've hit similar roadblocks in trying to get the boot floppy/prom/NFS stuff to work. Too green in Linux so far...

I'd be very interested in your boot rom images. What ROM chip do we need? There's a burner at work... might be able to do these if I knew which one.

THANKS!!!!

 

[cory]

well here is a chunk of a /etc/dhcpd.conf file
that sets the ip address and downloads the kernel using tftp
it is for a local subnet.
after it downloads the kernel
it mounts the root files system from
/tftpboot/192.168.100.10 (or whatever the ip address is)
on the main linux box.
in that nodes root filesystem you could also have it mount a shared
/usr filesystem or any other filesystems (see below for fstab)

i do have a little script that i found that sets up
the /tftpboot/192.168.100.10/ dir's for each node with
just what is needed

hope this will help some

this is /etc/dhcpd.conf file
subnet 192.168.100.0 netmask 255.255.255.0 {
option subnet-mask 255.255.255.0;
option time-offset -6; # Eastern Standard Time
host node0 {
hardware ethernet 00:50:ba:85:a3:75;
fixed-address 192.168.100.10;
filename "/tftpboot/vmlinux";
} host node1 {
hardware ethernet 00:50:ba:85:9d:74;
fixed-address 192.168.100.11;
filename "/tftpboot/vmlinux";
}
}
this is part of /tftpboot/192.168.100.10/etc/fstab file for
mounting root and /usr filesystem

192.168.100.1:/tftpboot/192.168.100.10 / nfs
rsize=8192,wsize=8192,timeo=14,intr
192.168.100.1:/usr /usr nfs rsize=8192,wsize=8192,timeo=14,intr
~

 

[Skaven]

Also in search of a Linux boot disk with the RC5 client!!

Lots of computers that are not setup with an OS.

If somebody can help me find something that would work.

All computers are x86, with 3com 3C905x nics..

Where shall I start? I'm new at Linux, but I got my RH6.0 system to share my cable connection at home.. so I kinda know my way around.

Thanks!

-Skaven

 

[Batti]

I sent an email off to a d.net board member asking about the single-floppy client he has, and why ccalvert's email hasn't been answered. I hope it generates some assistance!

 

[Batti]

Out of curiosity, what's the beef with using ccalvert's floppy? As long as he's using the official client, isn't it legit for him/us to create the OS as we need?

P.S. - just installed a Celly 566a and overclocked it to 850 to help make sure we make it Saturday night.

 

[Russ]

GOT IT!!!!! Hah! No stupid node is going to beat me!!!! Node 1 is now crunching/fetching/flushing like a good little boy! Beat that sumbitch in to submission!!! Now, on to his brothers. Bwuahahahahaha!!!!!

Russ, NCNE

 

[Batti]

Uhh, Russ, you are still working, aren't you? Hate to see you lose the shop to a crack addiction...

 

[Michael]

What are you changing, btw, to make it work.

Michael

ps - Russ'll expand his business from this. He'll be able to do custom Linux installs soon. No OS cost for the PC sale as well.

 

[Russ]

Batti,

Not to worry. There are 24 hours in a day; I use a lot more of them than most people. Besides, this time of the year things are always slow, so I have a little extra time.

Michael,

You're sort of right. Not really planning on Linux systems, but I am using this as an opportunity to grow my network and connectivity knowledge. I get an ever increasing number of requests from clients in this area and it would be nice to have answers and help for them.

What I did to get it working; first, I figured out how to disable to built-in firewall in the Linux distro. My hunch is that it was using NAT and since my DSL router is already using it, they may have been stepping on each other.

BUT, that didn't solve it completely. I than went in to the client options and told it NOT to grab a keyserver automatically, and hardwired one of dnet's keyserver IP addresses and, bingo! she works.

Russ, NCNE

 

[Batti]

Russ,

You are correct about FreeSCO using NAT out of the box. I've had problems with having a FreeSCO router behind a full linux box router in the past - I suspect the firewall rules don't like that much.

You don't need NAT in the scenario you are using...

 

[Russ]

Okay, all five slave nodes are up and cracking. Add 2002Mhz of power to TA Cube. Total current rack power, 3102Mhz. Time to build more nodes!!!!

Russ, NCNE

 

[BoberFett]

NAT by itself wouldn't be a problem, you could have as many gateways/firewalls on a network as you want. However multiple machines running DHCP on a network tend to disagree with each other.

 

[Russ]

BoberFett,

Yep, that was one of the many challenges I ran in to along the way. Freesco also has two different settings for DHCP; one to tell if to act as a server, the other to have the NIC grab and address FROM a server. Beat on that one for a while before I figured out that it shouldn't be doing both.

Russ, NCNE

 

[Kilowatt]

Not to rain on your parade Russ, but you really should put a Dnet PProxy on one of your 24/7 machines connected to the Internet.
If Dnet blacks out again for a little bit, or your DSL line goes kuput, your rack might end up doing randoms.

I,ve had the cable modem go out for a couple of days before, and almost ran out of blocks.
Now I hold 20,000 in the PPRoxy, and flush/fetch the PPRoxy every 2500 blocks.
I'm averaging about 6000 a day, so it will supply enough for at least 3 days worth.

It really is alot easier to track which node might not be responding, or went down, by looking at one .log file in the PProxy.

Besides, then you can do a stats deal just for your rack, and (almost) instantly see which node is working.

Either way, I'm impressed with your desire to get this rack up and going.

That's My 2 cents worth.

 

[Russ]

At this stage in my progession, pproxy is way over my head. Definitely something for later consideration, though.

Russ, NCNE

 

[BoberFett]

If you're not ready to run a Linux pproxy, you can always run one on a Windows box. It's extremely easy to setup and there are a number of people here that would gladly help you set one up, myself included. For the number of machines you have going I'd definitely recommend a pproxy.

 

[osmo]

I agree with BoberFett on the Windows PProxy,

I have one here at the office which is only a dozen comps or so. After a couple of DNET blackout's, it's worth the extra setup. Which BTW is easy!

Osmo.