[295] in Moira
Re: comment/question on previous item in discuss meeting
daemon@ATHENA.MIT.EDU (dkk@ATHENA.MIT.EDU)
Sat May 4 14:17:04 1991
From: dkk@ATHENA.MIT.EDU
Date: Sat, 4 May 91 14:16:14 -0400
To: haynes@cats.UCSC.EDU
Cc: bug-moira@ATHENA.MIT.EDU
In-Reply-To: haynes@cats.UCSC.EDU's message of Sat, 04 May 91 10:15:05 PDT,
> What other things (besides syslib, default printer, and Zephyr
> server) might be appropriate to put in the cluster services?
If you implement a per-cluster message-of-the-day, as has been
considered here (many times...), you would probably want to specify
the motd-server separately for each cluster, to distribute
administrative control. (So one cluster's administrator can change
the message on a Saturday, rather than wait for a central
administrator to come in on Monday morning.)
If you do load-sharing or batch-processing, you might want to use
clusterinfo to partition that, too.
Assuming split administrative control, you may want to have
cluster-administrator fields which tell your home-grown
problem-reporting programs (somewhat like Athena's "sendbug") where to
send mail when hardware breaks, and such.
If you put too much in the clusterinfo, you might run into the UDP
packet size limit. One solution is to have a small tree of Hesiod
info. Here's a simple example from an MIT/Athena workstation:
------------
tisiphone% hesinfo m66-080-3 cluster
lpr homer
syslib m66-vssys
tisiphone% hesinfo m66-vssys filsys
RVD vssys CALLIOPE.MIT.EDU r /srvd
RVD vssys POLYHYMNIA.MIT.EDU r /srvd
------------
The "syslib" field points to an fsgroup so there's a fallback in case
the primary system file server is unavailable.
You may want to do something like this:
------------
cats% hesinfo dogs cluster
lpr milkbone
syslib pets-sunsys
cmotd VET.UCSC.EDU
admin pet-owners
cats% hesinfo pet-owners addrlist
hardware operations@cats.ucsc.edu
software sysadmins@cats.ucsc.edu
management theboss@lion.ucsc.edu
------------
("cmotd" == cluster-motd)
This assumes you're willing to add the new Hesiod types to Moira, or
just maintain them manually on the Hesiod servers.
Warning: I have never implemented these suggestions, so I may be
overlooking something obvious.
