5 to 6

[Andrew Reilly]

On Fri, Oct 20, 2006 at 12:56:58AM +0100, Robert Watson wrote:
> On Thu, 19 Oct 2006, Randy Bush wrote:
> >do folk actually successfully upgrade
> >to RELENG_6 *safely* on a many-user production system using the 
> >instructions in UPDATING?

I've done that successfully on two single-user workstations, but it
failed when it came to try on the one multi-user production box
that I maintain.  I've posted to the lists about this a couple of
times and received zero responses, so my guess is that my condition
was sufficiently obscure that no-one else has experienced the
problem[*].  A shame.  I'll have to do a re-format and start from
scratch upgrade, some time that I can get the system quiet enough
for long enough.

> Do make sure to boot a 6.x kernel with your 5.x userland and test things 
> out a bit - make sure your hardware probes, is reliable, etc.

Ditto.  This was how I discovered that the ataraid in the 6.x
kernel does something sufficiently different to the one in the 5.x
kernel that I couldn't get past the fsck of the /usr partition...
Luckily enough of / was mounting successfully for me to back out to
the (working) 5.x kernel.

So: my two cents: it can work, but it's possible for it not to
work, and care is required.

[*] The production server is using a software RAID mirror on
a pair of SATA drives on a low-end Intel P4/ICH6 motherboard,
using ar(4), configured by atacontrol.  Fsck on 6.x can't find
any superblocks on /usr, but 5.5 is fine.

Cheers,

-- 
Andrew