After Wednesday’s triumph with the hardware, today I elected to fight the battle of the software, and take the OS to 10.5 (Leopard).
As with the hardware upgrade, the first order of business was cloning the boot drive using Carbon Copy Cloner. Just like Wednesday, this took a couple of hours, and appeared to work just fine. I popped the Leopard disc into the Mini, and let the machine reboot off the install disc. Everything appeared normal… until the installation stopped, telling me I couldn’t install to the boot drive. Panic stricken, my brain leaped to all kinds of conclusions — failed drive, lost data, and all kinds of awful things. As it ends up, somewhere along the way, the new 250GB drive I installed the other night used an Apple Partitioning Scheme rather than a GUID Partition Table. First off, I’d never known these were even options, and I certainly didn’t know how the 250GB internal was set up with the Apple scheme. In any case, that was a Bad Thing — the Leopard upgrade wouldn’t go until the drive was re-partitioned with the GUID Partition Table setup. That meant everything would be lost on the 250GB, leaving the USB connected 120GB as the (potential) master copy. Urgh.
Remember the nice 2GB freeware disk image from OWC? My guess is that when they partitioned the drive to put that on there, they used a bit more “universal” partitioning scheme. Not a big deal, but that’s one thing I’ll remember to look for in the future before trying these kinds of hijinks!
I started looking to see if you could boot a Mini from an external USB drive, and found you could, although almost everything I read indicated that Firewire would work better, and that the USB-booted OS would … run … very … slow. Figuring the only real path I had here was to boot from the USB drive, repartition the 250GB drive, and then recopy from the 120GB to the 250GB, I set about getting that going. The system did boot from the USB drive, and it really wasn’t that bad. A little slow, but since all I was doing was copying the data across, there wasn’t any real processing going on, so no big impact on the copy time.
Knowing that my sites had been down for several hours already, I decided I needed to set up a “Gremlins Are on the Loose” page to handle requests, knowing that the Mini would be down for at least three or four more hours. After a little bit of searching for the location of the system-wide default page (/Library/Webserver/Documents, just in case you’re looking for it), I dropped it into place on the MacPro, and pointed all HTTP traffic to it from the router.
So two hours later, it was time to reboot from the newly GUID’d 250GB drive, and make sure that Tiger was working. The boot seemed fine, so it was off to try the Leopard installation again. Boot the disc and… no GUID issues! Now, it was time to let Leopard do its upgrade voodoo. What’s odd to me is that the DVD media, once you’ve selected all your options, does a self-health check to ensure that its consistency is good before installing. While I appreciate the paranoia, I also appreciate the “Skip” button for this step, and skipped the DVD health check. So almost four hours after I thought I was ready to upgrade, I was actually upgrading. Now I just had to wait another hour or so for the OS to be upgraded!
Unfortunately, after the upgrade, neither my webserver nor mailserver were up and running. 🙁 Well, not running right anyway.
Leopard takes Apache from a 1.x codebase to 2.x, and in the process, changed the directories for users’ individual webserver configuration files. These moved from /etc/httpd/users to /etc/Apache2/users. I’d remembered the files, but forgotten the location and naming convention for them, so they were a bit of a bear to find and scoot from the old structure to the new. Once I did, Apache started, but would bomb on any page that was requested from it. The culprit here was a commented out php5 directive in the httpd.conf file for Apache 2.x. Not hard to change, but was a bit surprised that php wasn’t turned on by default. A quick uncomment, restart, and things appeared to be good on the webserver front.
The mailserver was a little trickier, but the messages in the mail log gave me the answers. For some reason, the UID used for the mail server changed. Not quite sure why Leopard did that, but a little change to a config file and a restart of the mail server, and all was well… I could see the spam floodgates opening! 🙂
However, I still couldn’t send or receive mail from my MacPro.
This one was a little harder to fix. Apparently, xinetd has been dropped with Leopard, in favor of launchd. Well, I’ve never used launchd, so that was a total mystery for me. Thanks to my good friend Google, I was able to find an example plist for my pop server, and that allowed me to get the pop server back up again.
So, from all appearances, things are back to normal at the Deauxmayne, nearly twelve hours (with breaks for dinner, visiting the gallery, etc.) after I began what I thought would be a small update. If something appears to be squirrely, let me know!