On Jan 26, 2007, at 1:14 PM, Tom Duerbusch wrote:
It doesn't make any difference as long as you and who ever else
ends up
in charge, knows what they are doing and when it happens.
I just want to point out that SNA makes available, free of charge, a
little utility I wrote called SYSVINIT, which is designed to a) not
pollute your SYSTEM CONFIG, b) deal with the idempotence issue of
AUTOLOG1, c) let you define individual services, with dependencies,
and base runlevels on those services, d) give you an interface to
start and stop services, and check their status from an authorized
userid with the TELL command.
...well, that's weird. I *thought* it was on the web site.
If anyone wants a copy of SYSVINIT, write to me off-list. You run it
in the AUTOLOG1 userid (although you could have a different userid
you started from AUTOLOG1 and run it there, if you wanted). It
basically just uses PROP to accept various commands and start/stop/
check services, and you can easily write Rexx to define new services.
My autolog1 takes 30 minutes to finish. I put it to sleep at several
points in the execution. The first one, is after SFS and VTAM are
brought up. It gives me a 2 minute window to see that we didn't
come up
right (controllers not powered up after a power outage...that sort of
thing) in that I can reipl without a lengthly shutdown process. Then
comes my critical and production machines, wait 15 minutes, my test
machines, wait 10 minutes and then "auditor" to pick up any minor
service machines. Obviously, in my case, the grants would be prior to
the first sleep period.
If you were using SYSVINIT you wouldn't need to hardcode the waits;
you could create dependency groups and check events, and stall until
the check (which is polled periodically, and you choose the period)
succeeded.
Adam
