Here's an interesting problem I want to share with the *DSM community
at large.   ( We are on AIX 4.3 with ADSM 3.1.2.20. (until this Sun.)

This morning I was prevented from doing any ADMIN commands to check up on
my system.   I use the admin command line client most of the time.

After a long while of failing to get in and do a query proc, I remembered
to do a tail on the adsmlog/output.log1 to see the last thing happening.

Incredibley it was up to date! & to the minute! Indicating that the system
was still running.   I then attempted a restore of 1 file and that worked!

So on thinking about it, I realized I caused the problem.  I was watching
my morning script run and decided I wanted to tweak it.   I do all my
script changes with the WEB browser interface.  It hung on the save and
I didn't think twice about it but just killed my browser.

It seems to me that ADSM runs a script directly out of where it resides in
the DB rather than making a copy.  BAD BAD BAD
 (children: can you say:  "Broken As Designed?")

SO I called IBM and asked if there's a way to 'signal' adsm to do the
halt command for me without logging in.  And they said being on AIX I
could try a:

     kill -15  (dsmserv's pid)

And it worked!  I.E> the log file shows normal shutdown procedures as when
you do a 'halt.'   and everything came back up just fine afterwards.

So moral:  don't change running scripts and remember kill -15 for Unix
*SM's if you can't sign in!!

           cheers!
                     .... joe.f.



Joseph A Faracchio,  Systems Programmer, UC Berkeley

Reply via email to