SUMMARY: Suggestions for a safe way to automatically do maintenance on remote machines ?

From: Bill Hart (hart@ml.csiro.au)
Date: Thu Apr 25 1991 - 12:48:30 CDT


My original question was :

        "........can somebody point out to me a simple elegant and
reasonably safe method for automatically doing such things as installing
patches and other simple tasks which involve mindless drudgery."

        I recieved lots of requests for summaries from people in a similar
position to myself (administrating part time and having enough machines to
make it a real chore to have to log into each machine to change something).
There seems to be two strategies for management in this sort of situation,
crisis-management (leave things alone until a crisis occurs, you do the
work your'e suppossed to do in between the crises), or to try and effectively
manage (set things up logically, reorganise when necessary, avoid continual
quick fixes, and bully youre users into not doing things in a non-standard
way). I've done both, both seem to consume the same amounts of time, but
effective management is less stressful, much more esthetically satisfying
and things do work better (it is difficult to explain to your supervisor
why it is necesary to spend a week or so setting up automount NIS etc, when
as far as he's concerned everything works ok, and we've only got two months
to finish an environmental impact study).

        Anyway sun-managers is not really the place for a discussion of the
ethics, aesthetics and job satisfaction of network management (my apologies,
although perhaps there should be room for discussion of strategy). I was
hoping for some tips from the "big guys" who manage hundreds of suns, but
what is effective for 200 machines may not be for 10 or 20.

        Saying all that I did get one positive response from
jam@philabs.philips.com, who does basically what i was thinking of doing :-
>
> We do things in a similar vein. We use the automounter with certain
> maps standard. One of the things we require and install as we set the
> machine up is an administration directory (/usr/share/adm). It is
> automounted, but doesn't have to be, it could just as easily reside in
> /etc/fstab. Under this directory is an etc directory (where we can put
> a standard printcap, aliases file, and the like) and a script
> directory that we require cron run the daily, weekly, and monthly scripts.
>
> Works quite well as we can put in changes and be assured that each machine
> will run them that night. You sometimes have to be careful to write your
> scripts in such a was that they won't do bad things if run twice or be very
> sure that you remove it after one night and the like.
>
Some mechanism for only running the scripts that haven't been run before
would be nice (we have machines that may go away to sea for a while) so it
wouldn't matter if a machine was down, and some system of reporting problems
would also be good. Security is also another consideration.

So I'll try to set up such a system myself over the next month or so and
summarize my findings to those who requested one. If anybody has anything
to add feel free to email me.

Thanks also to bb@math.ufl.edu who pointed out a possible cause and solution to
my problem of machines hanging.

And although I sometimes curse the flood of non relevant (to me) stuff that
comes flooding out of sun-managers, it has saved me much time and heart ache.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Bill Hart Internet : hart@ocean.ml.csiro.au
CSIRO Division of Oceanography Phone : +61 02 206 446
Hobart, Tas., 7000 Australia Fax : +61 02 240 530
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:13 CDT