What kills ypserv?

ypserv runs on 2 different "yp master" nodes
for our 29-node farm. On one of the masters
(actually, the "backup" master, really) ypserv
tends to suddenly stop working, which further results
in communication breakdowns between the nodes and the
server where ypserv quits, which in turn cause data prociessing
jobs that - among other things - move data back and forth between
the nodes and the ypserv-less server, to hang and fail.
My questions are:

1. What sorts of things can cause ypserv to die? (I am hoping that
I can find out by running it in debug mode, but I won't be surprised if
I can't.)

2. Why don't the nodes pick up the other, primary master as their master
node? This is supposed to happen automatically, as I understand it.

Thanks in advance,
Joe


joe

On Fri, 14 Jan 2005 01:36:56 +0000, joe wrote:

> ypserv runs on 2 different "yp master" nodes
> for our 29-node farm. On one of the masters
> (actually, the "backup" master, really) ypserv
> tends to suddenly stop working, which further results
> in communication breakdowns between the nodes and the
> server where ypserv quits, which in turn cause data prociessing
> jobs that - among other things - move data back and forth between
> the nodes and the ypserv-less server, to hang and fail.
> My questions are:
>
> 1. What sorts of things can cause ypserv to die? (I am hoping that
> I can find out by running it in debug mode, but I won't be surprised if
> I can't.)

Dunno. Bug? "fat fingers"?

> 2. Why don't the nodes pick up the other, primary master as their master
> node? This is supposed to happen automatically, as I understand it.

It is supposed to, and it does on my SuSE (8.2 and now 9.1) systems, as
well as my Solaris systems. I do recall seeing some "weirdness" between
ypserv on Linux and Solaris, I think it was related to defaults and DNS
fallbacks, but I don't remember the details. My workaround was to use the
Solaris systems for NIS and main NFS servers. Linux should still work tho.

What distro are you running? for workstation? for server(s)?

--
Juhan Leemet
Logicognosis, Inc.