Even though the filer can have multiple network interfaces, it does not function as a router; that is, the filer does not route packets between its interfaces on behalf of other network hosts. It can, however, route its own packets.
For routing its own packets, the filer relies on the default route and explicit routes. Typically, the filer learns explicit routes through icmp redirect messages received from the default router; you do not need to enter explicit routes in the filer's routing table. To display the filer's current routing table, use the netstat -r command. For example:
netstat -r
Routing tables
Internet:
Destination Gateway Flags Refs Use Interface
default nwo- UG 1 138 e0
298.295.227 link#1 UC 0 0 e0
298.295.227.1 link#1 UHL 1 24 e0
nwo- 8:0:20:79:f9:79 UHL 1 0 e0
298.295.227.255 link#1 UHL 1 3696 e0
If you must enter explicit information into the filer's routing table, use the route command. See the na_route(1) man page about how to add or modify information in the routing table.
If the filer cannot find an explicit route in the routing table for a particular destination, it uses the default route. This means that the filer sends the traffic to the default router, which is specified in the /etc/dgateways file in the root volume.
One default router is specified during setup. You can, however, add potential default routers at any time to the /etc/dgateways file in the root volume. For each added router, you should also specify the metric, which is a number indicating the route preference for the router. The highest preference is 1, which is also the default preference for the router specified during setup. The lowest preference is 15. For information about the format of the /etc/dgateways file, refer to "The /etc/dgateways file."
To help manage multiple routers and to enable you to create redundant routing schemes, the filer runs the routed daemon, a simple routing daemon, which is enabled at boot time. This daemon "listens" for Routing Information Protocol (RIP) packets being exchanged between routers on the network to determine which routers are alive. From the routers that are alive, the routed daemon selects the one with the highest preference to use as the filer's default router.
However, the filer doesn't rely on the routed daemon to construct the routing table. The function of the filer's routed daemon is to check the status of the default router. Refer to the na_route(1), na_routed(1), and na_dgateways(1) man pages to learn more about routing on the filer.
You turn the routed daemon Off and On with the routed command.
To turn the routed daemon Off, enter
routed off
To turn the routed daemon back On, enter
routed on
Note: If you turn the routed daemon Off by editing /etc/rc in the root volume, manually designate a default router in /etc/rc.
To display the status of the default gateway list, use the routed status command. The -n option forces the command to display numeric values for gateway names. An example of the routed status display is as follows:
routed status
RIP snooping is on
Gateway Metric State Time Last Heard
karl 1 ALIVE Wed Mar 9 03:38:41 GMT 1994
groucho 1 ALIVE Wed Mar 9 03:38:41 GMT 1994
292.0.266.366 1 ALIVE Wed Mar 9 03:38:41 GMT 1994
292.0.266.377 1 ALIVE Wed Mar 9 03:38:41 GMT 1994
june 2 ALIVE Wed Mar 9 03:38:41 GMT 1994
292.0.266.332 2 ALIVE Wed Mar 9 03:38:41 GMT 1994
292.0.266.352 3 ALIVE Wed Mar 9 03:38:41 GMT 1994
292.0.266.351 4 ALIVE Wed Mar 9 03:38:41 GMT 1994
292.0.266.350 5 ALIVE Wed Mar 9 03:38:41 GMT 1994
The routed status display shows the following information:
The /etc/dgateways file in the root volume is the configuration database for the routed daemon. From the routers that routed has determined to be alive, routed selects the one with the highest preference to be the default router. When the filer cannot find an explicit route for a packet, it routes the packet to the default router.
The file consists of lines with the following format:
gateway metric
where gateway is the name or the IP address of a default router and metric is a preference indicator, which ranges from 1 (highest) to 15 (lowest).
| #Gateway | Metric |
|---|---|
| 192.9.200.10 | 1 |
| eng_gateway | 2 |
Note: Each entry for such a default router must have an IP address that belongs to the IP subnet of one of the interfaces configured for the filer.
The following list describes how the filer uses its interfaces to respond to different types of packets.
The filer does not use the conventional IP routing mechanisms to reply to NFS-over-UDP requests. The filer sends the response on the network interface on which the request was received to the same address that generated the request. For example, the filer named toaster uses the toaster-e1 interface to send packets in response to NFS requests received on the toaster-e1 interface.
This way of handling NFS-over-UDP requests enables you to attach multiple interfaces of the filer to networks with the same IP subnetwork number while keeping NFS-over-UDP traffic isolated to the appropriate physical networks.
Because of this scheme, it is possible that NFS-over-UDP responses might be returned through a different path than you might expect from an examination of the IP routing table using netstat -r. This scheme generally works well, although it can result in different routes than expected if your environment contains one-way routes. For example, the IP packets might not be routed as you intended if you configured the network so that the IP traffic from host1 to host2 is routed through router1 and the IP traffic from host2 to host1 is routed through router2.
The filer tries to return NFS-over-TCP and -HTTP traffic over the interface on which the traffic was received. However, there are exceptions. For example, if the filer experiences excessive queuing to a response or experiences a time-out followed by a retransmit, the filer routes the traffic by using conventional IP routing table lookups. If the filer has multiple interfaces attached to networks with the same IP network number, the filer uses the first interface it finds with that number to send the responses.
Note: If the NFS-over-UDP, -TCP, -CIFS, or -HTTP traffic takes place on an ATM interface, the filer always uses conventional IP routing table lookups and routing for that traffic.
For other types of traffic, for example, traffic generated by telnet, rsh, and ping, the filer uses IP routing table lookups and routing. If the filer has multiple interfaces attached to networks with the same IP network number, the filer uses the first interface it finds with that number to send the responses.
Copyright © 1999, Network Appliance, Inc. All rights reserved.
Part No. 210-02334
File: net4.htm
Last updated: 04/26/99 08:11:18