Kicking and screaming…
I remember looking at LiveUpgrade at my old job and it not working very well because I was trying to be all tricksy with it and make it do things beyond the most basic “clone this and apply these patches” — I was trying to use it to convert Solaris 8+VxVM to Solaris 10+SVM, and while it should have worked, it didn’t, and it felt like the failure to work was down to arbitrary limits and hard-coded inanity in LU.
So I’ve never really seen it working in any useful fashion.
This week, while thinking about how best to set up the new shinies (an X4600 to play the role of database server, and a pair of X4100s to be its minions) LU somehow popped into my mind. As I have a little time to play with for a change, figured it couldn’t hurt to fiddle.
So we now have a new default configuration for all new Solaris systems. Single filesystem for the OS at ~50GB, a spare slice of the same size for the alternate BE, with the rest going to swap. This is assuming 136GB spindles.
And it works quite nicely. There’s some risk of actually being able to patch things in future, thanks to the easy back-out route and the much less-insane downtime requirements.
I’d like to think that we’ll migrate existing systems to a similar configuration, it’s certainly technically possible to do, but I’m not going to count on having the time to do it. Shall have to settle for a very gradual process of conversion by replacement.
Popularity: 24% [?]