I fixed the NFS problem on image that Charlie noticed. It turns out that nfs-utils links against the kernel, and something got munged when I downgraded the kernel. I rebuilt nfs-utils from source, and it works now.
I've got Jabber configured on image. I just have to work out some kinks wrt authentication.
I've gotten the Digi driver and software installed on Image.
For future reference, the files and documentation can be found at http://www.digi.com/support/productdetl.jsp?pid=2253&osvid=103 This is the download site for Fedora Core 2 packages, although there are packages for numerous Unices, minus *BSD.
I downloaded the kernel source from Red Hat because the documentation claimed to need it, and I followed the instructions about cleaning the version files (although the would have already been clean since it was a new download). So I guess they were just linking against it.
I installed using the SRPM since this is an RPM based distro. After doing the rpmbuild in /usr/src, the RPM it built was placed in /usr/src/redhat/RPMS/i386. I then installed this with rpm.
I could go on longer, but I'd just be reading the documentation.
I've gotten MRTG running on image, and am monitoring both the Cisco routers and both the Extreme switches. Go to mrtg.cs.earlham.edu to see the graphs.
I've moved image's Ultra3 SCSI card into quark, to accomdate our plan (hope? dream?) of having our tape backup gear hook directly up to quark. All appears to be good. Kudos to Matt Hogan in helping me get quark in and out of the rack.
Here is the procedure for doing an image dump to quarkprime's DDS-III drive:
1. Label a fresh DDS-III tape with the current month.
2. Pop it in quarkprime's DDS-III drive.
3. Run this command on quarkprime as root:
RSH=/usr/bin/ssh dump 0aun -f 192.168.1.2:/dev/sa0 /var/lib/systemimager/
I've successfully moved Nagios from millie to image, leaving millie available for cluster work. I installed all the plugins from RPMs, and then copied all the configuration stuff to image. I made sure to change the paths where necessary.
I've restored FAH on image from backup. It appeared to have gotten lost in the shuffle.
Image is now working. Apparently rsync requires a group with the ID 'root' to have rsync work, so I created that and updateclient is now working on the ACLs.
I modified Quark's dhcpd.conf so that option-100 points to the new image's IP address. I then imaged ACL0. It ran into a problem in not being able to find /usr/share/systemimager/boot, which it needs to boot off of, so I scp'ed the files over from the new image. The image then went smoothly, except for one error:
2004/06/25 13:53:46 [5096] rsync on scripts/hosts from acl0.cs.earlham.edu (159.28.230.10)
2004/06/25 13:53:46 [5096] link_stat "hosts" (in scripts) failed: No such file or directory
2004/06/25 13:53:46 [5096] rsync error: some files could not be transferred (code 23) at main.c(397)
I don't have time to look into this further, but I'm guessing that this is a configuration problem on the new image.
Well, I installed Fedora Core 2 on Quarkprime this morning as the first step in our transition of swapping Q' and image. I did an NFS install off of office0, and it was fast and painless. The entire installation was pretty much a breeze. Some things I should point out about the install:
The aftermath:
Systemimager
While I was upgrading everything else, I thought I might as well upgrade image, to keep things consistent. It now has OpenSSL 0.9.7d and OpenSSH 3.8p1.
I've upgraded image, backup, and proto to kernel 2.4.25. No problems detected thus far.
image's system drive failed today, probably due to spindle failure. I've swapped an identical drive, and have installed Red Hat 9. Hassan has installed the C3 tools, SystemImager and setup passwordless authentication to the ACL machines. I have also installed another 18GB IBM UltraStar, and have setup software RAID 0 between the previous UltraStar and the new UltraStar. There are some cooling issues that still need to be resolved before the machine can be moved back into the rack.