>I am in charge of a IBM pseries 680 Model S85, could somebody point me
>in the direction of recommended maintenance schedules for the machine
>e.g. power downs. Alternatively what recommendations can people make,
>does the machine require a reboot after ??? days. The machine is in a
>controlled enviroment eg humidity, ups etc etc and is running AIX
>4.3.3 with a financial application.
We've had machines go 700-900+ days at a time between reboots, no problem.
We could easily do *much* more, but we really don't like to go more than
120 days between reboots for reasons to be explained below.
Generally speaking, UNIX machines don't require a 'periodic' reboot. I
can't speak about non-UNIX/VMS/mainframe machines, however. ;)
Only reasons why you reboot an AIX machine is:
a) special upgrade involving kernel or critical libraries
b) device driver upgrade that can't take effect immediately
c) serious DoS attack situation where reboot is fastest way to recover
No reason to reboot a machine, otherwise. So we don't have a planned
downtime thing except for periodic OS/app patching -- we try to do this
every 3-4 months or so, in order to get the greatest amount of support
from IBM when we do call in a potential bug or * situation. For example,
sometimes IBM support will (understandably) insist that you patch your
system to at least the latest revs or even the most current maintenance
level patch set, and then try to reproduce the problem before they will
work further on it.
It also gives us time to test a new cut of a patch set on various hardware
and software combinations before committing it to production machines.
Financial apps generally means change management and testing is paramount,
so sounds like you'd probably be looking at a minimum of 6 (or even 12)
months between significant updates or reboots. Won't be a problem at all.
It all boils down to... you reboot your machines when you need to. ;)
When will you need to? (See items 'a', 'b', and 'c')
How often do you do these items (well, at least 'a' or 'b') will depend
on how much system/app downtime you are comfortable with and can afford.
Some sites, once every 20 years (like Irish Railways and their VMS system
that ran the track signal controllers)... some, once a week... some, once
every few months... some, once every six or twelve months, etc.
-Dan
(email sent to original poster as a courtesy and posted to comp.unix.aix)