Forum Moderators: bakedjake
Any comments on the upgrade process? I should be able to just drive to the datacenter, run an update/install using a current version of mandriva, restart the server and be good to go. However I'm suspicious that my experience will be as rosy as that :). Thinks like drivers, php and apache upgrades, and so on, seems like I could run into problems.
I don't have good experiences with Linux upgrades. I once upgraded from Debian 3.0 to Debian 3.1 which brought me into problems because some things wouldn't compile anymore with the new gcc version 3.x, where the older compiler 2.95 didn't like some of the header files that were also updated. I was stuck somewhere in the middle.
I don't know about your hosting location, but my hosting company once offered my a temporary server for free for a period of a month to do this type of parallel upgrade.
- Make sure your /home is on a different partition to / (I normally put all my website and sql data into /home)
- Backup
- Install new operating system (make sure you do not format /home)
- Restore anything not in /home
- You should be up and running
Quite often hosting companies reimage the server so you will have to backup /home to another server.
If you can suffer a few hours of downtime then it is easy, otherwise you will have to move the site to another box whilst you work on the main one.
Distros like Ubuntu claim to be able to update a live system, I have had limited success with this though. A deb based distro is better than an rpm based one, Gentoo is by far the most reliable to upgrade major versions but that is not on servers very often.
- Make sure your /home is on a different partition to / (I normally put all my website and sql data into /home)
- Backup
- Install new operating system (make sure you do not format /home)
- Restore anything not in /home
- You should be up and running
This doesn't account for problems with different Apache and MySQL versions, crontabs that start periodic processes, firewalls and everything OS related.
Different Linux versions may have different kernel versions with different needs, other modules to load or modules that have to be recompiled to run under the new version. One of my servers running CentOS 5 crashed recently when doing nothing more than updating a kernel with bugfixes from version 2.6.18-53 to 2.6.18-53.1.14. Altough this was a bugfix update only and the kernel version number remained the same, the RAID controller driver failed to load under the new kernel and caused effectively disk corruption. You don't want to have these problems on a production server.
There is only one safe way: second server->install->test->replace
You guys are full of cheery news aren't you :).
That's what I've done in the past, generally when I move data centers every 2-3 years for some reason. I'll put a new machine in the location, get it up and running, change DNS and then slowly decommission the old machine. I was hoping for an easier method I guess.
Thanks for the advice!