Hi,

I know this is a Linux forum, but i was hoping that someone might know where i can go or who i can speak to for MPICH help. i browsed the web for an MPICH forum, but the most recent postings i have found are 2000 from comp.parallel.mpi. so i assumed it's not being used anymore. but what do i know?

just in case you know anything about MPICH, here is my problem.

I'm currently trying to install MPICH on Linux RH7.3 (on a single PC.) I read somewhere that i can install MPICH even if i'm not using multiprocessors or a cluster. i just have to somehow trick the installation that it's installing on a single node with multiple processors.

this is what i did:
after downloading,

% ./configure --with -device=ch_p4 --prefix=/usr/local/mpich-1.3.4 > c.log

configtn completed except it couldn't find an acceptable F90 compiler or JAVA, but i don't think this will stop me from successfully installing MPICH. next i did:

% make > make.log
% cd examples/basic
% make cpi
% ../../bin/mpirun -np 4 cpi

this is where i got errors:
p0_2301: p4_error: Timeout in making connection to remote process on localhost.localdomain: 0
../../bin/mpirun: line 1: 2301 Broken pipe /usr/local/mpich-1.2.4/examples/basic/cpi -p4pg /usr/local/mpich-1.2.4/examples/basic/PI2175 -p4wd /usr/local/mpich-1.2.4/examples/basic

so according to the installation guide, "connection refused" may be caused by the Internet security settings on my system that restrict the # and frequency of interprocess connection operations and that i should run the following commands:

% iptables --list
% ipchains --list

and look for limits, restrictions on source or destination ports and such.

but when i ran % iptables --list i got the following errs:
/lib/modules/2.4.18-3/kernel/net/ipv4/netfilter/ip_tables.o: init_module: Device or resource busy
Hint: insmod errors can be caused by incorrect module parameters, including invalid IO or IRQ parameters
/lib/modules/2.4.18-3/kernel/net/ipv4/netfilter/ip_tables.o: insmod /lib/modules/2.4.18-3/kernel/net/ipv4/netfilter/ip_tables.o failed
/lib/modules/2.4.18-3/kernel/net/ipv4/netfilter/ip_tables.o: insmod ip_tables failed
iptables v1.2.5: can't initialize iptables table `filter': iptables who? (do you need to insmod?)
Perhaps iptables or your kernel needs to be upgraded.

the guide also says to check that /etc/inetd.conf allows more processes per minute for rsh. first, i don't really understand what that means and second i couldn't even find the file in /etc.

so this is where i'm stuck.

any help or references would be GREATLY appreciated. thanks

MPICH is one of those sticky areas, but I'll see if my limited experience can help any. First of all, MPICH is pretty advanced stuff. If you don't absolutely need it (ie. the only way to run your code is with mpirun in parallel or you can't compile a serial version for your uniprocessor machine) then I would stay away. OK, enough disclaimer.
You are correct, you won't need f90 or java, so let's see...The first step is to make sure you have rsh enabled. MPICH starts each new thread of the program using rsh -n, for /dev/null if I remember correctly. Most firewall set-up tools will automatically disable remote shell logins, even for ssh. Go to a prompt and type "rsh localhost" and if it tells you the connection was refused then you'll know you need to turn it on. About iptables and ipchains - if you aren't using either of those as a firewall tool, then trying to list the rules won't get you anywhere. If you don't have an inetd.conf file, you'll proably have an xinetd.conf file. They're pretty much the same thing and that's the one you need to edit. Read the man page for xinetd.conf and it explains what it does and how all of those lines work. The other thing that you'll need to make sure you have is a machines.linux file. I think this will be in the util directory MPICH intalled. In there you should list localhost several times, one for each "processor" you want to use. I think that by default the installation program will list it five times. The one final idea I have is that you might want to use the ch_shmem instead of ch_p4. Not sure about this, but since you're running this all on a local machine with the same memory and hard disc resources shared memory may be the way to go.
Hope this helps. Let us know your progress.

Good Luck,
Senor Pedo

hi senorPedo (is this meant to be spanish? - just curious)

i have some questions, and i just wanted to excuse myself in advance because my questions may be trivial

first, i typed
% rsh localhost
and again i was told connection refused

so you told me to enable rsh, but i don't how to do that. so i was wondering if you could give me some guide?

next, how would i know if I am using iptables or ipchains as a firewall tool? oh let me tell you that I am working on a linux that is installed as a dual boot system at work. and i know there is a firewall, but that's all i know, i don't know any specifications on it. also, i don't know if it matters, but basically everyone else has Windows NT, so i dont' know how that affects anything.

i also checked my xinetd.conf file and from reading around, it seems that it is fine. it is as follows:

defaults
{
instances = 60
log_type = SYSLOG authpriv
log_on_success = HOST PID
log_on_failure = HOST
cps = 25 30
}

includedir /etc/xinetd.d

so i don't understand if any edits need to be made if any.

the reason i installed ch_p4 is b/c eventually we want to purchase a beowulf cluster, and i thought it would be a good idea to imitate that environment. i don't if there are any differences in the programming technique, but i'm assuming there are.

and just to let you know why i'm doing all this, i'm an intern and i was given a serial program in F77 to parallelize and eventually give some #s on how the speedup is for a certain # of processors. i think i know how i would program it (it's pretty easily parallelizable and it is a program that is computationally intensive), but i need a way to start programming it in parallel and test it. so that's why i was looking for some mpi environment. i know i won't get the actual speedup on a single cpu but at least if i have mpich installed i could start learning how to program in mpi and i could have a final program that is proven to run in parallel so that i can just use it on the beowulf once we get it. if you know of any tools i can use that will analyze the speedup given a certain # of processors that can run on my current machine, i would also greatly appreciate that too.

again, i'm sorry for all the trivial questions.

OK, ch_p4 is the correct choice for your application. You'll have fun getting this all set up.

Anyway, To turn on rsh, you'll actually go into the directory /etc/xinetd.d and there will be a list of services that can be run on the computer. The three that you'll need to make sure are on (by editing the file to make sure it says disable = no) are rsh, rlogin, and rexec. After enabling these, type /etc/rc.d/init.d/xinetd restart and try to log in from your console. Type rsh hostname (localhost or whatever your computer is named) and you should be able to log in. There may be one more file you have to modify to allow for the null login...you'll have to give me a day or so to think on that one.

Don't worry about the firewall stuff, if it's not actually on your computer itself (which it sounds like its not) then it doesn't matter. As for any sort of benchmarking or scaling testing, I'm not familiar with any tools in particular, but they're out there. Try googling for some. I do know that there are tons of scientific papers floating around out there on the scalability of large clusters.

Best of luck in getting this all up and running! It will be fun once it's all together. BTW-someone once told me that Senor Pedo literally translated means Mr. Fart. So I figured hey, why not. Any other questions you have I'll be happy to try and answer. And your questions so far have not been trivial-this is tough stuff. It's been a long time since I played with getting MPICH set up but once you have the basics it's easy to run with it.

Good luck!

thnx for replying,

so i was about to make changes to rsh, rlogin and rexec, until i realized that they don't even exist in /etc/xinetd.d

so what i did is i started to search for examples of these files, this is what i created:

% emacs rsh

# default: on
# description: The rshd server is the server for the rcmd(3) routine and, \
# consequently, for the rsh(1) program. The server provides \
# remote execution facilities with authentication based on \
# privileged port numbers from trusted hosts.
service shell
{
disable = no
socket_type = stream
wait = no
user = root
log_on_success += USERID
log_on_failure += USERID
server = /usr/sbin/in.rshd
}

% emacs rlogin

# description: rlogind is the server for the rlogin(1) program. The server \
# provides a remote login facility with authentication based on \
# privileged port numbers from trusted hosts.
service login
{
disable = no
socket_type = stream
wait = no
user = root
log_on_success += USERID
log_on_failure += USERID
server = /usr/sbin/in.rlogind
}

% emacs rexec

# description: rexecd is the server for the rexec(1) program. The server \
# provides a remote login facility with authentication based on \
# privileged port numbers from trusted hosts.
service login
{
disable = no
socket_type = stream
wait = no
user = root
log_on_success += USERID
log_on_failure += USERID
server = /usr/sbin/in.rexecd
}

then i typed
% /etc/rc.d/init.d/xinetd restart
% rsh localhost

it still told me connection refused

while i was searching for the rsh, rlogin and rexec files i read somewhere that i should edit etc/sysconfig/ipchains
and this is what i have now for ipchains:

# Firewall configuration written by lokkit
# Manual customization of this file is not recommended.
# Note: ifup-post will punch the current nameservers through the
# firewall; such entries will *not* be listed here.
:input ACCEPT
:forward ACCEPT
utput ACCEPT
-A input -s 0/0 67:68 -d 0/0 67:68 -p udp -i eth0 -j ACCEPT
-A input -s 0/0 67:68 -d 0/0 67:68 -p udp -i eth1 -j ACCEPT
-A input -s 0/0 -d 0/0 -i lo -j ACCEPT
#
# New rules for rlogin/rsh traffic, incoming or outgoing
#
-A input -p tcp -s 0/0 -d 0/0 513 -b -j ACCEPT
-A input -p tcp -s 0/0 -d 0/0 514 -b -j ACCEPT
#
# End of new rules
#
-A input -p tcp -s 0/0 -d 0/0 0:1023 -y -j REJECT
-A input -p tcp -s 0/0 -d 0/0 2049 -y -j REJECT
-A input -p udp -s 0/0 -d 0/0 0:1023 -j REJECT
-A input -p udp -s 0/0 -d 0/0 2049 -j REJECT
-A input -p tcp -s 0/0 -d 0/0 6000:6009 -y -j REJECT
-A input -p tcp -s 0/0 -d 0/0 7100 -y -j REJECT

i also read that i should edit /etc/hosts.equiv but it doesn't exist either. so i created the file and it holds:

localhost.localdomain

so after making these modifications i restarted xinetd and rsh localhost, but i still got connection refused. i don't know what i'm doing wrong.

oh one more thing, i checked if i have /usr/sbin/in.rshd, /usr/sbin/in.rlogind and /usr/sbin/in.rexecd but i don't. what should write for these??? thank u so much

btw, yes, Senor Pedo does mean Mr. Fart in spanish.