Finally
It’s only been a year since I wrote http://personal.rwcinc.net/archives/2006/05/24/server-upgrade-postoned/, but it’s time to upgrade.
(Note: I had more on the front page, but I didn’t get everything I wrote copied onto the site :()
Not wanting to have the server down for too long I started making some changes on the old machine to get all of my configurations copied/backed up. I run a regular backup process on a nightly basis that copies all of my configuration files and is useful if I had a catastrophic hard drive or similar failure that allowed me to rebuild the machine - it wasn’t really designed to allow this upgrade process. My initial plan was to move the old drive containing /usr, /var and /home and to copy the configurations that I needed back onto the server. That turned out to not be as simple as I had thought. It would have been better (and I’m planning on some “major” configuration changes to make this easier the next time - cause there will be a next time) to have the important files linked in instead of trying to mount the entire partition - I won’t bore you with the details, suffice it to say that I will be making changes to the /var and /usr directories over time.
I figured that I might be able to have the server down for somewhere between 20 minutes and an hour. I was so very wrong. The server closet is really a closet. Amongst the clothes, an unused bicycle and a set of luggage sit two servers. One is mine, the other is Scott’s. The roof slopes down to two feet high on the far end, a lone bare bulb burns on the far wall as my only light source. The small wicker stool is just high enough not to cramp the legs, but I spent most of my time bent forward reading the monitor that sits on the floor. When I thought I had made all of the changes/backups that I would need I opened up the case of the new box. It has slide out trays on the front for easy hard drive changes which are nice, since I was beginning to worry that I would need adapter rails for the swapped over hard drive. I then opened the side of the old machine so as to have as little downtime as possible. The Intel boxes that I got from Steve have these nice removable cages that the drives sit in. They hold a floppy and a couple of hard drives and are held into the chassis with a single screw and a tab; remove the screw, push the tab and they slide right out. I shut down the machine. As quickly as I can I pop out the cage and unscrew the hard drive, moving it over to the new server. And this is where it all starts.
- June 12
After sliding the new tray into place, I discover that the IDE cable plugged into the CD Rom drive is not long enough to plug into both the hard drive and the CD Rom. I have an extra cable, but this motherboard only supports one IDE cable, unlike most computers that I’ve worked on. It makes sense, of course, since it’s designed to support SCSI drives and even has two different SCSI controllers for different types of SCSI drives. Since I didn’t have a large enough cable for both drives (I should get cables from Jake) I merely unplugged the CD drive and plugged in the hard drive.
Five hours later, after managing to mess up partitions (hint never, never, ever link to an fstab file on another partition. Ever!), unplugging the hard drive and plugging in the CD drive to boot into rescue mode, copying, linking and reconfiguring I was back online - at least partially. Everything looked good from where I was sitting, so I asked my daughter if she was hungry - she’d been hanging out with me the whole time I was there - and we went to a late lunch.
After I got back home, I tried to log in to get my email. Damn! Something was wrong, I couldn’t get my email. No problem, just log in and see what the problem is. Oh man. SSH is configured wrong. Hop back in the car, cross town, copy the right configuration file and restart SSH. I check it from another machine onsite. Yes, I can log in. I drive back home to login remotely to fix the email issue. Oops. There was a service not started. The defaults are crazy… at least to me. Most everything is turned off. It really should be so that one has to consciously start them. OK good.
- June 13 AM. I discover that I can’t send email. I’m supposed to be on a trip to the Grand Canyon today so it’s not getting fixed this morning.
- June 13, PM. I find another service that I didn’t start - after fumbling around trying to find why I can’t send email. I can send email again
- June 14, AM. My mailing lists are awefully quiet. Try a test message, the machine replies no such user. Arrrgh. Another configuration file that didn’t copy. Fixed.
Now that it’s all over I have a few notes on how I’m going to do this different/better next time.
In the meantime, this server seems to be quite snappy. It’s only an 800MHZ machine with 512MB RAM, but that’s a huge jump from 400MHZ/128MB.
