Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!rice!sun-spots-request
From: barmar@think.com (Barry Margolin)
Newsgroups: comp.sys.sun
Subject: Re: YP fault tolerance.....
Keywords: Networks
Message-ID: <6092@brazos.Rice.edu>
Date: 26 Mar 90 05:54:58 GMT
Sender: root@rice.edu
Organization: Sun-Spots
Lines: 18
Approved: Sun-Spots@rice.edu
X-Refs:  Original: v9n97
X-Sun-Spots-Digest: Volume 9, Issue 97, message 10

The original poster asked about YP slave servers "taking over" when the
master server dies.  YP servers don't keep track of each other at all.  It
is the clients' responsibility to do this.  As far as clients are
concerned there are no differences between master and slave servers.  When
a YP request times out the client sends out a broadcast looking for a new
YP server.  The first server to respond (which can be either a master or a
slave) is then used for future queries.

Since broadcasts only work within a single subnet, the most important
consideration for YP fault tolerance is that there must be at least two YP
servers on each subnet at sites that have multiple subnets.  If there's
only one server on your subnet and it goes down you'll never be able to
find an alternate server automatically (but ypset can be used to manually
bind to a server on another subnet if you have done "ypbind -ypset").

Barry Margolin, Thinking Machines Corp.
barmar@think.com
{uunet,harvard}!think!barmar