On 5/31/12 8:13 PM, Matthew Seaman wrote: > On 31/05/2012 16:41, Damien Fleuriot wrote: >> You missed the bit about 3 reboots, while these don't take 15 mins each, >> they're still time consuming and disruptive. >> 1/ reboot after installing new kernel >> 2/ reboot after installing new world >> 3/ reboot after rebuilding ports > > If you rebuilt the ports first, then you'ld only have two reboots. > > Also, while the cautious approach detailed in /usr/src/UPDATING is never > wrong, much of the time you can do the upgrade perfectly well by > installing world+kernel together and just rebooting once. Obviously > this is not a good idea if your machines are in a datacenter many miles > away and you don't have console-equivalent access or if you're upgrading > over a large delta in versions, or you're making major changes to the > kernel config. > > This sort of operation is something that ZFS boot environment support > (recently committed to HEAD, due for MFC within the month) makes much, > much safer and easier to deal with. You don't need to do a separate > reboot to test the kernel as you've still got an entire kernel+world in > the previous BE to fall back on. > > Cheers, > > Matthew >
The reason I rebuild the ports last is because, unless I'm wrong, any port that's statically linked to a system library would be linked to the old library from the old world. We've got very high HA constraints on these machines and I really prefer doing this the cautious way. Hell, on the first reboot I actually test the new kernel with "nextboot -k" , even when doing 8.2-RELEASE -> 8-STABLE upgrades... Regarding the ZFS boot thingy, I'm not comfortable enough with it to push it in production, so we're still using UFS here. Sure looks interesting though. _______________________________________________ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to "freebsd-stable-unsubscr...@freebsd.org"
