This page servers as a place to discuss changes to node configuration, and running the cluster. There is a separate wiki page for the software itself.
The magus cluster is currently running on several machines donated by Eastern Michigan University's Computer Science department. The nodes are currently running one of three branches, although only 0.2 has been active lately. (0.1.1, 0.2 and CURRENT)
- CURRENT nodes need to be updated to a recent version. However, some changes with the network stack need to be worked out first.
- 0.2 nodes will need to be updated after the /bin/sh fixes have been MFC'd
- We need to generate new chroot tarballs on i386 for 0.2-PRERELEASE. I suspect this will break GNUstep ports.
- For nodes at EMU, we should consider switching to NFS instead of scp. The machines are physically close to the file store.
- We should consider setting up a proxy server just for nodes. That might minimize fetching distfiles over and over again on different nodes. It looks like magus keeps distfiles on the same node per run.
- Nodes now use our default firewall rules if they're running 0.2. We need to make adjustments as some sites don't allow fetching distfiles with this setup.
- We need to enable cpufreq on the other nodes. During a run, the CPU stays at 100%, but power has become a concern when they're idle. Most 0.2-PRERELEASE nodes have it enabled.