Virtual test cluster
We have the following nodes (all kvm virtualized). You should be able to access them from the sheltered network.
beta - 192.168.12.59
beta000 - 192.168.12.60
beta001 - 192.168.12.61
- etc (up to 003)03)
beta is the master.
The nodes do not have individual disks, the root filesystem on the nodes is shared at beta:/srv/beta-root. It is the same filesystem for _all_ nodes. Hostname and IP address is set via DHCP, you can use conditionals on the hostname if you need to make scripts aware of where they are running. Use
schroot # as your own user (not too useful) schroot -u root # as root
to run commands in the chroot. Be aware that not everything is mounted in the chroot
If you run apt-get in the chroot, it won't start any daemons. Don't leave programs running in the chroot.
Since it is a shared filesystem, you can't hard code any names or IP address in the config files. You have to get the config files to automatically detect that stuff. If you need hard-coding, I have several ways to work around it.
ssh as your user, use sudo. Add people to group sudo to give them sudo permission.
Planning new cluster
- Have it share uids/gids with Cal's cluster, so that we can cross-mount /home (we can even use cal's home export for now...)
- What box will we use for the NFS server/DHCP/NIS, etc, for our cluster?
- We need a private subnet with our own DHCP server in order for this to work - we can get that allocated from Cal, say 192.168.15.0/24.