Migration planning - old system to new
John
john at starfire.mn.org
Sat Jan 23 17:08:44 UTC 2010
On Sat, Jan 23, 2010 at 11:19:34AM -0500, Jerry McAllister wrote:
> On Sat, Jan 23, 2010 at 10:15:19AM +0800, Erich Dollansky wrote:
>
> > Hi,
> >
> > On 23 January 2010 am 01:12:19 John wrote:
> > > Now that I've actually gotten the new system to boot, I need to
> > > figure out how I'm going to migrate everything - users, data,
> > > MySQL, NAT, firewall, apache, DHCP, gateway services BIND,
> > > Sendmail, etc., etc from
> > > FreeBSD 4.3-RELEASE #0: Thu Jan 22 19:44:16 CST 2004
> > > to
> > > FreeBSD 8.0-RELEASE #0: Sat Nov 21 15:48:17 UTC 2009
> >
> > this is real jump.
> > >
> > > Bit of a challenge, eh?
> >
> > I have heard that somebody actually landed on the moon? Was it
> > you?
> > >
> > > Not only that, but I'd like to update my UID scheme from a
> > > pre-standard version (most of the UIDs are down in the 100s) to
> > > the new convention so that I'm more in-line with the rest of
> > > the world.
> >
> > Ok, I cannot imagine how you will do this with the access rights
> > of the files?
> > >
> > > My rough idea:
> > >
> > > 1) Create a "migrate" account in Wheel with home as
> > > /var/migrate so that I can do a dump/restore on "home" without
> > > messing things up
> >
> > Are you sure? Use /usr to make sure you will have enough space.
>
> You are making the rash and probably incorrect assumption that /usr is
> the largest partition/filesystem. Many people, including I, make /home
> or another partition be the large one. The OP may also have done that.
>
> >
> > > 2) Start putting together all the pieces - trying to find
> > > update / conversion scripts whenever possible.
> >
> > I think, this would only help if you would go the long way 5.x,
> > 6.x, 7x and finally 8.
> >
> > Setup the new machine, install the applications you need,
> > configure them as close as possible to the original configuration
> > and see what happens.
> >
> > > 4) Let people move in, try it out, see how things are
> > > 5) Fix everything found in #4
> > > 6) Try a cut-over and make sure all the network services work
> > > in the middle of the night sometime, then switch back
> >
> > Oh, it is a life system in use while you migrate.
> >
> > Are you able to set the new thing up in parallel?
> >
> > It might be easier for you to run both machines and move first the
> > simple things over.
> >
> > > 7) Nuke /home and /var/mail and migrate them again to get the
> > > latest version 8) Do the real switch
>
> Move/migrate them first. Don't make assumptions about what the OP has
> on /home.
>
> But, I agree, if possible, use a second machine with V 8.0 installed
> and migrate to it.
>
> Otherwise, make full backups, check them for readability. Then do a new
> install of FreeBSD V8. Add a large disk and pull stuff out of your dump
> to it and then migrate that stuff piece by piece back to the machine
> main filesystems.
>
> ////jerry
>
>
> > > 9) spend a couple of weeks fixing all the things that weren't
> > > so disastrous that they got picked up in #4.
> >
> > I think, if you do it service by service, you have a better chance
> > to avoid this.
> > >
> > > Ideas / scripts / project plans / outlines - whatever? Maybe I
> > > should write a chapter for The Complete FreeBSD after surviving
> > > this...
> >
> > Yes. It is a Le Must.
> >
> > Erich
Sorry, gang - I should have been more clear! I am DEFINITELY doing
this on a new machine! And I don't need any "migration" storage,
because, well, gosh - it's tcp, people! ;) I just did the first
transfer of home, and it went swell:
On elwood (the new, 8.0 system):
cd /
umount /home
newfs /dev/ad0s3e
mount /home
cd /home
rsh dexter "dump 0uf - /home" | restore rvf -
That preserves all the file modification times, too, even on directories.
All you young'uns out there - don't forget dump/restore! tar and cpio
are nice, but sometimes, you just gotta take it all...
("dexter" is the old machine - the FreeBSD 4.3 system)
Oh, BTW, just for giggles:
10:56AM up 492 days, 13:57, 2 users, load averages: 0.02, 0.03, 0.00
That's right! Nearly 500 days! And it was well over a two hundred
days before that, but we had a power outage that outlasted the UPS.
Gotta love this stuff!
Name Mtu Network Address Ipkts Ierrs Opkts Oerrs Coll
rl0 1500 <Link#1> 00:40:f4:8e:37:c7 384602213 1 585428419 0 0
rl0 1500 192.168.1 gateway 56160554 - 40918536 - -
rl0 1500 dexter/32 dexter 246146 - 0 - -
ed0 1500 <Link#2> 52:54:40:21:f8:a8 609350703 376495 380734152 0 15678720
ed0 1500 XXmaskedXX xxMASKEDx x 63482137 - 380606442 - -
lo0 16384 <Link#3> 27069210 0 27069210 0 0
lo0 16384 127 localhost 27069192 - 27069192 - -
987522505 cpu context switches
2590296969 device interrupts
342786031 software interrupts
1096125243 traps
1217705867 system calls
4 kernel threads created
12006286 fork() calls
429899 vfork() calls
0 rfork() calls
954 swap pager pageins
1156 swap pager pages paged in
456 swap pager pageouts
774 swap pager pages paged out
7135 vnode pager pageins
32897 vnode pager pages paged in
1 vnode pager pageouts
1 vnode pager pages paged out
1502 page daemon wakeups
2926806 pages examined by the page daemon
7262 pages reactivated
507170456 copy-on-write faults
0 copy-on-write optimized faults
175434998 zero fill pages zeroed
127442529 zero fill pages prezeroed
1028 intransit blocking page faults
1020317711 total VM faults taken
0 pages affected by kernel thread creation
1300974057 pages affected by fork()
49625615 pages affected by vfork()
0 pages affected by rfork()
808730759 pages freed
4 pages freed by daemon
578473126 pages freed by exiting processes
10408 pages active
41102 pages inactive
2502 pages in VM cache
7573 pages wired down
2148 pages free
4096 bytes per page
637017826 total name lookups
cache hits (69% pos + 4% neg) system 6% per-directory
deletions 0%, falsehits 0%, toolong 0%
--
John Lind
john at starfire.MN.ORG
More information about the freebsd-questions
mailing list