Configuring a Relay Agent

It is likely that most or all of your NFS cluster's clients are attached to the same subnetwork as your cluster server. If not, a gateway (a device, such as a router or computer) can be used to connect two or more networks.

The figure below shows a cluster server with two LAN cards connecting the server to two subnetworks (LAN1 and LAN2). Cluster clients can be put on one or both of these subnetworks. These clients are local clients (not remote). The figure also shows a router that connects LAN2 with LAN3; any cluster client that is put on LAN3 will be a remote client.

Server

+----------------+ | |

| LAN Card ===----------------------+ |

| | | |

| LAN Card ===----------------+ | |

+----------------+ | | |

| | |

| +--<Router>--+

| | |

LAN1 LAN2 LAN3

Once a gateway is attached, the server can boot clients that are on subnetworks that the server is not directly attached to. There can only be one gateway that separates the server from the remote client. That is, you could not add a router that connects LAN3 with yet another subnetwork and put clients on that subnetwork (more details about this restriction is given below).

A relay system is a computer that is on the same subnetwork as the clients to be booted. A relay agent is software on the relay system and server that is configured to pass client and server messages between the two subnetworks.

There are some restrictions in setting up a relay system:

The relay system must be a Series 700 or Series 800 computer in the same subnet as the client. This machine must be running HP-UX 10.01 (or later) from a local file system; that is, it cannot itself be a client of another NFS cluster.

The client must be only one hop from the server; that is, the client and server subnetworks must be connected through a single router or gateway. You can verify this by running /usr/sbin/ping with the -o option from the relay system to the server. For example, to check the hops from tinkrbel to peter:

tinkrbel: /usr/sbin/ping -o peter -n 1
PING peter.neverlnd.com: 64 byte packets
64 bytes from 153.13.115.149: icmp_seq=0. time=18. ms
----peter.neverlnd.com PING Statistics----
1 packets transmitted, 1 packets received, 0% packet loss
round-trip (ms)  min/avg/max = 18/18/18
1 packets sent via:
      153.13.112.1     - croc-gw.neverlnd.com
      153.13.115.149   - peter.neverlnd.com
      153.13.104.1     - croc-gw.neverlnd.com
      153.13.105.109   - tinkrbel.neverlnd.com

Note that the packet went from the relay system tinkrbel via the gateway croc-gw to the server peter and returned back to tinkrbel via croc-gw. This shows that tinkrbel is only one gateway (croc-gw) away from peter.

To configure the relay agent, follow these steps:




	NOTE: You must make the changes on the relay system manually (that is, without using SAM). Later, when you use SAM to configure a gateway client, use the IP address of the relay system in the "`Default Route`" field of the "`Define Clients`" screen.

In the file /etc/inetd.conf, add the following line if it does not already exist:
bootps dgram udp wait root /usr/lbin/bootpd bootpd
In the file /etc/bootptab, add the following information for each client that may be booted across the gateway served by this relay system. (See bootpd(1M) and comments in the file /etc/bootptab for further information.)
# The following is a client that boots from server:
client's_host_name:\
ht=ethernet:\
ha=client's_hardware_address:\
bp=server's_IP_address:\
hp=1
The hop count, hp, must be 1.
For example, using the information displayed by the ping command above to configure client wendy to boot from server peter across a gateway using relay system tinkrbel, install the following entry on tinkrbel:
# client 'wendy' (ha) boots from server `peter' (bp)
wendy:\
ht=ethernet:\
bp=153.13.115.149:\
ha=08009935c990:\
bp=153.13.115.149:\
hp=1
In the /etc/rc.config.d/netdaemons file, set the value of START_RBOOTD to 1 to ensure that the rbootd daemon starts at boot time:
START_RBOOTD=1
If it is not running already, start the rbootd daemon (see rbootd(1M)).
The rbootd daemon provides NFS diskless cluster support for Series 700 clients with older boot ROMs designed for the DUX clustered environment without requiring boot ROM modifications (SAM automatically configures rbootd on the cluster server).

Naming services used by the server are not transferred to diskless clients that boot over a gateway. If the server uses DNS or NIS services, these services will have to be manually configured on the gateway client. Gateway clients are only provided with a copy of the server /etc/hosts file.

Configuring a Relay Agent

Technical documentation

» Table of Contents

» Index