From: jreed@appliedtheory.com
Date: Mon Aug 05 2002 - 12:07:58 EDT
I asked about an issue where quorum seemed to exist in a 2-node memory
channel cluster with a quorum disk, but clu_quorum gave errors:
framework error: connection problems
Mike Caplin came back first and hit it on the nail, because he'd been there:
--------------------
"cfgmgr is needed (in /etc/inetd.conf) in order for clu_quorum and
clu_upgrade to work properly."
--------------------
We typically remove almost everything from /etc/inetd.conf for security
reasons, in fact on Suns we are considering not running it at all. However,
on the cluster you need, of course, the rshd, and turns out you need cfgmgr.
Dr. Tom Blinn comments at some length, appended below. He talks about
ics subsystem, but I didn't find any ics* entries I could identify in
our saved inetd.conf original file, so I don't think it's a problem as of yet.
Here are his comments:
-------------
"The "clu_quorum" stuff uses, I believe, the ICS subsystem (the
intracluster or internode communication services), which has all
those "ics" daemons that provide the services. If you removed
any of the "ics" stuff from your /etc/inetd.conf or otherwise
have messed up the ICS definitions, you may have broken things.
The complaint about the "framework error: connection problems"
sounds remarkably like what I'd expect to come out when there
are problems communicating between the nodes using the normal
communications framework that needs to be there for the cluster
to work."
--------------
This would explain all the ics* processes one sees on the cluster, but
it appears they do not start from inetd.conf, I don't know where they
*do* start from.
Thanks to those who replied so quickly, I added cfgmgr back in and
all is well!
Judith Reed
This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:48:48 EDT