Urgent help needed: Licensing snafu following server crash

Fairlight fairlite at fairlite.com
Fri Sep 21 10:27:28 PDT 2007


You'll never BELIEVE what Barry Wiseman said here...:
> I do see that the yum daemon installed a new kernel (CentOS 4.5,
> 2.6.9-55.0.2 -> 2.6.9-55.0.6) on Sep 4.  The post-crash boot would be the
> first instance of the box running that new kernel.  Might fP's licensing
> scheme have stumbled over that?

My rule of thumb: *nix should -not- be auto-updated by an auto-scheduler.
You just found out precisely why, and my experiences with SuSE's SP1 boot
loader menu writing glitches are also an example (along with a few other
oddities I've seen people run up against over the years).

You might use, say, yast or up2date or whatever is on your system, but
start it manually, under controlled circumstances, at a time of your
choosing, when you can afford to take the time and roll things back,
troubleshoot, etc.  And reboot after kernel patches when you can afford at
least an hour downtime, giving yourself enough time to revert.

It may work for 'doze, and I auto-update everything in 'doze, including my
apps.  But I -never- use auto-update scheduling programs in *nix.  Ever.
I've seen it cause way too many debacles.  I honestly hold the opinion that
*nix systems should be attended upgrade events where one can devote one's
full attention and resources to them, even if they're supposed to take 5min
and go off without a hitch.  I also hold the opinion that unixen should
be administered by real admins, while 'doze boxes really don't require
that extensive a level of knowledge.  I'd say something like Server 2003
requires some extra, obviously, but still nowhere near the intricacy.

YMMV, but your recent experience should be enough to illustrate I may be
onto something.

mark->


More information about the Filepro-list mailing list