Re: [OT] suites (was Re: re evolution)
On Sat, Apr 15, 2006 at 07:14:07AM +0200, Francesco Pietra wrote:
> Could you please show again the recipe (parallel operation) for carrying out
> upgrades? i missed it and i am not sure to have the right one. please
> indicate if 64 is comprised. thank a lot
> fp
My system is was set up with one partition containing almost everything,
with only /tmp and /home being separate partitions.
I create a new partition (this does assume you have the disk space
available) and copy the main partition of my running Debian system into
the new one. Then I adjust the boot system (I use Lilo) so that I can
boot from eirher copy. To do this, I also have to adjust /etc/fstab of
the new system so that it will use the new system as / and not the old
one.
Once I have a dual-boot system that will boot either one, and have
checked that they work, I make boot floppies for booting each of them,
just in case.
One of these systems is going to be upgraded; the other is the spare.
I make sure that I use the lilo in the spare to make the entire system
bootable, whether from hard disk or floppy. Again, it doesn't hurt to
have yet one more extra floppy that uses the lilo in the upgradable
partion as well. I like to have as many independent ways of booting
as possible. Booting from Lilo actually uses some data on the system
you originally ran lilo from, so I make sure I have boot possibilities
from either copy of the system.
Now, once you are satisfied that either system will work properly, I
boot the upgradable copy and upgrade that.
If anything goes wrong, well, I can use the other one for production
use until the problems are ironed out.
There are a few potential glitches.
If you have /var, /usr, and so forth as separate partitions, you are
going to have to make copies fo them, too.
I usually don't make duplicate /home partitions, but share /home
between the two systems, so that any production work I do on one is
available on the other. This isn't foolproof, because is, say,
openoffice or major database system is unpgraded, it may convert my
files from an old format to a new one, and then I am unable to go
back. And mail queued in one system may be unavailable to the
other, so it's a case of elther living with it or messing around
with symbolic links, or a separate shared /var/mail partition.
Some versions of exim seem to refuse to deliver to a symbolic link,
so that cab get complicated, too.
And, just in case, I back up all the file systems that would end up
shared between the two systems and might suffer incompatible
changes. I believe in *lots* of backups. I'm th guy that often posts
on this list just to remind people to make backups.
-- hendrik
>
> On Friday 14 April 2006 16:42, you wrote:
> > On Wed, Apr 05, 2006 at 02:30:48PM +0100, Nikolay Kichukov wrote:
> > > Hi there,
> > > Thanks for sharing your way of doing upgrades. I will
> > > consider that idea in the future.
> >
> > It's called parallel operation, and it's what I've been advising
> > everyone to do when performing any big system change for about thirty
> > years now. And it's important to do it with the complete production
> > environment, not just a representative sample.
> >
> > -- hendrik
Reply to: