The first part of the day for me was spent working on looking into the SystemImager Auto-install kernel. This kernel needed recompiling so that it would work with the various cards that we are using in the ACL and the Athena Cluster. The later part of my day was spent "playing" with the wireless equipment that Aaron and I brought over from the Ranch. So-far, I'm having trouble establishing a high-quality connection from one of the PCMCIA wireless cards to the Aironet base stations. The problem seems to persist regardless of the antenna used. The baffling thing is the fact that I can establish an excellent connection to the Apple Airport with the same cards.
Well, I started out working with the wireless equipment again. No joy... I can't seem to get things to work properly with the Aironet bases. Well, enough of that for now. I got back to compiling the kernel for the SystemImager Auto-install disk. I managed to get the kernel configured properly so that it works with the 3c509, 3c59X, and eepro100 Ethernet drivers. I didn't get to actually doing the make dep, make clean, and make bzImage today, but that shouldn't take too long tomorrow. I also upgraded the version of SystemImager that we are using on the ACL to 1.5.0. My reasoning behind this was that the new version now includes support for SSH and also has a push update script for pushing updates to clients. This significantly cuts the time that I would have had to spend writing that script (Yeay!!!). The Image and the server now include the new version and it will be propagated to the running ACL boxes tonight at 3am.
Got the SystemImager Auto-install Kernel compiled and installed into the proper place (/tftpboot/kernel) so that it is put on the floppy the next time that I create a boot disk. Unfortunately it would appear that this has no effect on our problem: How to get the ISA Ethernet adapters to work with the auto-install kernel. I am truly baffled. I sent an email to the systemimager-discuss list hoping that they know more about kernel drivers and how the autinstall kernel is configured. The one response that I got back was to pass the kernel a irq setting in case the kernel was unable, for whatever reason, to autodetect the adapter. This didn't produce any change. I've given up on this approach for the time being since it is working on a problem that only arrises when we are re-installing the machines (which hopefully be very infrequently). Tomorrow I will work on finishing up the installation of the golden-client (athena0). With luck I can get it squared away and have the image ready for distribution on friday.
Hypercube routing is so interesting... In my networks class in AU we didn't get much into the theory behind it, but now after having Aaron showing it to me once, I really wish we had... Anyhow, thats what a good portion of today was spent doing. Aaron and I worked out the hypercube routing scheme (on paper at least). While we did that I re-ran the installation for Debian on athena0. Turns out that the last time that it got installed there were a few things missing. Regardless, I tried to split my time since the Install does require a significant amount of waiting. After that, I got the image pulled and tested it on athena1 and athena2. Sofar things seem to be working just fine. The only problem that I ran into was the fact that the DHCP client dameon kept re-writing the /etc/resolve.conf file telling it to look at localhost for the nameserver. I don't know how or where to fix this since the DHCP config on Noether seems to be fine and shouldn't be causing this problem. For now, I set-up the client (pump) to not rewrite the file as a short term fix.
TODAY WAS REALLY WORTH IT!!!!!! By the end of the day (midnight), I managed to get the entire PVM synchronized and running the way it should. Of course there are some crucial things that we need to finish before proceeding. Namely, although it is installed and compiled, the pvfs needs its last few adjustments so that it starts working correctly. This shouldn't take too long on Monday. One thing that hindered me in getting the cluster synchronized was what appears to be a defective floppy drive in athena12 (aka athenac). I will/have documented how I managed to synchronize the machine nonetheless in the SystemImager documentation. Currently there is no cron-tab syncing the machines every day so we will have to push any changes manually.