From saville at comcast.net Sat Aug 23 12:55:02 2003 From: saville at comcast.net (Gregg Germain) Date: Tue Nov 9 01:14:27 2010 Subject: [scyld-users] Help! Endless RARP requests Message-ID: <3F479E28.B0653915@comcast.net> Hi, I have the SCYLD Beowulf Basic edition (Red Hat 6.2) which I got from Linux Central. I installed it on a PC with 2 ethernet cards. The installation went gret. Next I connected another PC to the Master node by dirctly connecting the slave nodes eth0 card to the Master Node's eth1 card. I then booted the slave node using the SCYLD CD. Booting seems to go well but thren it gets to the point where it's making RARP requests and it stays there forever. Eventually the slave node reboots itself and then gets stuck inthe RARP request once again. Can anyone give me a hint as to what I need to do to get the master node to respond? an /sbin/ifconfig on the master node shows that the eth1 card is active and operating in the range selected at install time. Do I need to connect through a hub? I know I have to get on eventually for the second slave, but I thought I coudl get the first one going without a hub. thanks for any help yu can give. Gregg From velthuis at fisher.biology.rochester.edu Sun Aug 24 10:50:06 2003 From: velthuis at fisher.biology.rochester.edu (velthuis@fisher.biology.rochester.edu) Date: Tue Nov 9 01:14:27 2010 Subject: [scyld-users] RE: Help! Endless RARP requests In-Reply-To: <200308231903.h7NJ3uw09904@NewBlue.Scyld.com> Message-ID: Hello Gregg, Here are some suggestions... Are you using a cross-over cable to connect the two PC's? direct connections need a different cable than is used when connecting via hub or switch. If a node can not reach the master (or, more accurately can not reach the bpserv deamon running on the master) it defaults to rebooting and will try again. During installation on your master, which eth port did you specify for the Beowulf network? (Assuming you used defaults throughout you can check your network is up by doing e.g. a 'ifconfig' or a 'netstat -rn' and look for confirmation that ipaddress 192.168.1.1 is up, and that you can 'ping' it etc.) Ok, you wrote that this seems to works. Next, do a 'ps ax' on the master to check that the bpserv deamon is running as it should. Pay attention during startup of the master for any indication that perhaps something is going wrong that leads to this server deamon to not being started automatically. (A 'dmesg' or a 'cat /var/log/messages' may be useful to check startup messages if they went by too fast. Alternatively pressing 'shift/page up' during boot should let you browse back thru messages that went of the screen already) Similar for any indication that your network is actually up and running as you state it is. Are you convinced that your ethernet cards function properly? E.g. if you would install any operating system on these two machines, would you be able to 'ping' from one to the other and vice versa using your current cable? Here it would help to check the LEDs on both ethernet cards. Are they both flashing when your node tries to reach the master? Hope any of these tips will get you going again. Good luck! B-J. > Date: Sat, 23 Aug 2003 13:02:32 -0400 > From: Gregg Germain > To: scyld-users@beowulf.org > Subject: [scyld-users] Help! Endless RARP requests > > Hi, > > I have the SCYLD Beowulf Basic edition (Red Hat 6.2) which I got from > Linux Central. > > I installed it on a PC with 2 ethernet cards. > > The installation went gret. Next I connected another PC to the Master > node by dirctly connecting the slave nodes eth0 card to the Master > Node's eth1 card. > > I then booted the slave node using the SCYLD CD. Booting seems to go > well but thren it gets to the point where it's making RARP requests and > it stays there forever. Eventually the slave node reboots itself and > then gets stuck inthe RARP request once again. > > Can anyone give me a hint as to what I need to do to get the master > node to respond? > > an /sbin/ifconfig on the master node shows that the eth1 card is active > and operating in the range selected at install time. > > Do I need to connect through a hub? I know I have to get on eventually > for the second slave, but I thought I coudl get the first one going > without a hub. > > thanks for any help yu can give. > > Gregg > From gregg at head-cfa.harvard.edu Fri Aug 29 13:00:10 2003 From: gregg at head-cfa.harvard.edu (Gregg Germain) Date: Tue Nov 9 01:14:27 2010 Subject: [scyld-users] Cluster is up - how to program? Message-ID: <200308291643.h7TGhQQa015726@head-cfa.cfa.harvard.edu> Hi all! Thanks to some helpful folks I got my 4 nose Scyld-Beowulf cluster up and running (just needed to buy a hub). I've run sample programs such as mpi-mandel and it seems to work fine. Now I want to start writing programs of my own. I like to start with simple "hello world" type programs. However, I've looked on the Scyld CD and cannot find and tutorials or even sample small programs to work with. I know I have to lean MPI (doing that now ) but even when I get a simple program going (there's a simple hello world program in the MPI book I have), I don't now how to run it on a Scyld-based cluster. How do you get the Scyld system to take the program you've just built and deal it out among the slave nodes? Can anyone point me to a tutorial or a thread that discusses this? thanks for your help. --- Gregg "Improvise, adapt, overcome." gregg@head-cfa.harvard.edu Smithsonian Center for Astrophysics Phone: (617) 496-1558