# $Id: VMware-setup-summary.txt,v 1.8 2003/09/16 20:35:13 tjn Exp $ Instructions for use of VMware for OSCAR testing NOTES: * If you don't already have a copy of VMware you can grab a 30-day evaluation copy at http://www.vmware.com/. This has the same set of features but times out at the end of 30-days. You won't have to do things over if you do obtain a license after this initial install -- simply enter the registration code and it removes the 30-day timeout. * I use an older copy of VMware 3.1.1 (build-1790). I have not tried using the new 4.x series but assume it is similar. If you use the provided RedHat 9.0 tarball I have available from my test machine, you may have to convert things to work with 4.x ... I'm not sure? * I put everything in '/vmware' so the contents of all the tarballs will expect things to be in this location. e.g., linux-node1.tar.gz (9.8k) # Not sure why these node archives have linux-node2.targ.z (5.9k) # diff sizes, but they're almost the same linux-rh9.tar.gz (639M) * The compute nodes have their VirtualFloppy drives directed toward a boot-floppy image. By defalut they do not have the floppy mounted so they boot from the virtual HD, so to boot from a floopy, at VM startup click: Devices \---->floppy 0 \------>Connect I setup my test machine to have all boot-flpy image in '/vmware/flpy-imgs' and create a symlink to the appropriate image. The nodes look at '/vmware/nodeX-fd0' where X is the node's number, e.g., linux-node1, X=1. * You will need a pretty good amount of disk space on your test machine for this configuration of a "VMcluster". ~1.7GB Distro RPMS ~1.9GB Linux-RH9 VM (linux-rh9.tar.gz) ~600MB Compute node VM (each) (linux-node[1,2].tar.gz) * Lastly, the password for 'root' on the VM (linux-rh9) is "vmware". 0. Install VMware (i use an older version - 3.1.1) 1. Either install a new guest OS from scratch or use the available RedHat 9.0 tarball from my test machine. - Note, the uncompressed tarball takes approx 1.9G of space. 2. I place all my VM's in '/vmware'. This pathing is modifiable manually by editing the 'linux-rh9/linux.cfg' configuration file if needed (or by using the VM wizard). root# mkdir /vmware root# tar -zxf linux-rh9.tar.gz -C /vmware root# tar -zxf linux-node1.tar.gz -C /vmware root# tar -zxf linux-node2.tar.gz -C /vmware And add the boot floppy images, something like this should work, root# mkdir /vmware/flpy-imgs root# cp oscar-boot-flpy-2.3.img /vmware/flpy-imgs root# ln -s /vmware/flpy-imgs/oscar-boot-flpy-2.3.img /vmware/node0-fd0 root# ln -s /vmware/flpy-imgs/oscar-boot-flpy-2.3.img /vmware/node1-fd0 3. I typically create a 'vmusers' group to setup ownership/permissions for the VM and other files on the host system (ie. your native desktop). - For further details on VMcluster setup see the docs at http://www.csm.ornl.gov/~naughton/vmware.html root# chgrp -R vmusers /vmware 4. The RH9.0 VM will be the headnode for a "virtual cluster" with the nodes being simple VM's (ie. linux-node1.tar.gz & linux-node2.tar.gz). - The host machine (your desktop) should NFS export a directory with the distro's RPMS from all CDs -- needed by OSCAR for installs. root# mkdir -p /tftpboot/rh9.0_rpm root# cp -a /mnt/cdrom/RedHat/RPMS/*.rpm /tftpboot/rh9.0_rpm #repeat this for CD 1,2, & 3...yep it's a lot of space! (~1.7GB) root# cat /etc/exports # General 'share' export /tftpboot/share 192.168.152.0/255.255.255.0(rw,no_root_squash) #vmnet1 /tftpboot/share 172.16.233.0/255.255.255.0(rw,no_root_squash) #vmnet8 # RH9.0 /tftpboot/rh9.0_rpm 192.168.152.0/255.255.255.0(rw,no_root_squash) #vmnet1 /tftpboot/rh9.0_rpm 172.16.233.0/255.255.255.0(rw,no_root_squash) #vmnet8 5. Fire up the VM $ /usr/bin/vmware /vmware/linux-rh9/linux.cfg -x & 6. The VMheadnode will be multi-homed, ie. it has a private network (eth0) and one that has Internet access (eth1) via NAT and the host machine. - You will need to edit the '/etc/hosts' file to change 'nermal-1' to your host's desktop name. This value shows up in the VMheadnode's '/etc/fstab' file too... so edit it for NFS to come up correctly. - Note the linux-rh9 has a base almost stock install of RedHat 9.0 with the hard drive set to "Un-doable" mode. This is useful so that at the time you shutdown/reboot the "VMheadnode" you can simply click "Discard" and nothing will be written to the perminent disk and each test run will have a stock/clean RH9 install! ;) - The network & hostname were setup as well as two hack for dealing with RH9 on VM's as described in the "Hack/Fixes" area of this page, http://www.csm.ornl.gov/~naughton/vmware.html 7. Once the VMheadnode is up and running you should be able to connect to either the host filesystem - via NFS '/tftpboot/rpm' & '/tftpboot/share' - via SSH (I have my desktop in the /etc/hosts file as nermal-1 edit for your desktop) 8. I often have a CVS checkout of the OSCAR devel tree on my host machine's '/tftpboot/share' Which is available to all VMheadnodes over NFS. So when I Discard the HD data for the VM, I can start over without having to do a fresh CVS checkout. 9. At this point you're able to do testing for OSCAR however you choose: - a stable release from OSCAR.sf.net (e.g., v2.3.1) - a snapshot from CVS For further details on VMcluster setup see: + "VMware/VMcluster Stuffo" http://www.csm.ornl.gov/~naughton/vmware.html More specifically, "VMcluster Stuffo" (VMcluster-0.2.ps.gz) + SSS-OSCAR related items @ ORNL http://www.csm.ornl.gov/~naughton/sss-oscar + SSS-OSCAR http://sss-oscar.sf.net