This chapter describes what to do when your load-balancing network connection has problems. The chapter describes the following topics:
When you experience difficulty with the load-balancing network connection at a particular system, you can do the following:
Check the physical connections at the system; consult the owner's guides for the equipment.
Search or read the /var/adm/SYSLOG file and console window for error messages. If you find any Network Load Balancing Software driver messages, read about them in Appendix B, “Error Messages”.
Use lbstat -C to evaluate the state of the load-balancing device(s) on the system. In particular, the state for each Ethernet port attached to the load-balancing device should be LINK_GOOD. If a port does not show this status, check its cabling.
If the Network Load Balancing Software device is not configured for static trunking, make sure configured switch ports are not setup for load balancing, trunking, or link aggregation using the IEEE 802.3ad standard.
If the Network Load Balancing Software device has been configured for static trunking, make sure the switch has also been so configured and is not configured to use LACP (Link Aggregation Protocol) or 802.3ad.
If the Network Load Balancing Software connection has not been working since the last time the system was booted, or if this is an initial installation of a Network Load Balancing Software product, one or more of the following could be occurring:
The operating system has not been rebuilt to include the driver.
The Network Load Balancing Software interface or the aggregated Ethernets have not been configured properly.
Follow the instructions below to determine the cause of the problem:
Verify that the installed IRIX operating system and Network Load Balancing Software software are the correct versions by doing the following:
Determine the correct versions of IRIX and of the Network Load Balancing Software. The Network Load Balancing Software release notes indicate the correct IRIX and Network Load Balancing Software versions.
Use the versions command to display the installed release identifications:
% /usr/sbin/versions eoe.sw.base eoe1 date Execution Only Environment 1, 6.5
If the version is not correct, install the correct version.
Note: The eoe.sw.base subsystem must be 6.5 or later for the Network Load Balancing Software to work.
View /etc/config/nlb.options; use lbconfig to check for Ethernets attached to the load-balancing devices.
Use netstat, as shown below, to display the currently configured network interfaces.
% /usr/etc/netstat -in
Figure 4-1 shows example output.
If the Network Load Balancing Software interface is not displayed, continue to the next step. If the interface is displayed, but the configuration is incorrect, follow the instructions in Chapter 2, “Network Load Balancing Software Installation and Configuration” to reconfigure it.
Verify the Network Load Balancing Software entries in the /etc/config/netif.options file. For example, check the spelling of the network interface.
Use /etc/autoconfig to rebuild the operating system to include the Network Load Balancing Software driver. Reboot the system to start using the new operating system.
Invoke netstat -in again. If the Network Load Balancing Software interface is still missing, contact your service provider.
Follow these steps to resolve the problem:
Ensure that the system is using the operating system that was built most recently. Use /etc/autoconfig to rebuild the operating system, and then reboot to start using it.
Watch the messages on the terminal during restart to verify that each network interface is configured correctly. The messages should look similar to these examples:
Configuring lb0 as sys3 Configuring ec0 as gate-sys3
If the load-balancing driver is not mentioned on the terminal during startup, there is a problem with the software.
Use netstat to display the currently configured network interfaces.
% /usr/etc/netstat -in
If the interface is displayed, but the configuration is incorrect, follow the instructions in Chapter 2, “Network Load Balancing Software Installation and Configuration” to reconfigure it.
The /etc/config/netif.options file may have an incorrect entry (for example, a misspelled network interface); verify all file contents carefully.
If the interface does not display, the software may be dysfunctional. Contact your service provider.
If high packet loss is displayed by the ping command, follow these steps:
Confirm that the source and destination IP addresses are on the same subnet. If there was any doubt that they are, enter ping -R, or traceroute to check the route. (For information on the latter command, see its manpage, traceroute(1M) .)
If the source and destination IP addresses are attached to the same switch and ping still drops packets:
Enter netstat -C on the source and destination systems.
Enter ping -f on the source and destination systems to make the problem more obvious.
In the netstat output, inspect the packet counts and indications of input errors or output errors while ping is running.
If the output shows (approximately) one packet dropped per second, check for defective hardware. Make sure all cables are seated properly.
If the netstat output shows no such errors:
Consult the documentation for the switch and its management software.
Use statistics tools specific to the protocol you are using (IP, UDP, TCP/IP) to determine drops.
Check for defective hardware such as switch ports and problems with the cabling.
If ping does not elicit a response from any system, the /etc/hosts and /etc/config/netif.options files may not have been set up properly. For example, the files may be configuring the Network Load Balancing Software interface with the Ethernet IP address.
Verify that the IP (inet) addresses for all network interfaces are correct. To display the currently configured IP addresses, enter
% /usr/etc/netstat -in |
If the addresses are not correct, follow the instructions in Chapter 2, “Network Load Balancing Software Installation and Configuration” to reconfigure the Network Load Balancing Software.
If the problem persists, identify other problems, as described in “Verifying the Network Load Balancing Software Connection” in Chapter 2.
Silicon Graphics workstations and servers are capable of loading a small-sized version of the operating system (the miniroot) and booting themselves over the network; however, they are capable of doing this only over Ethernet local area networks (they cannot boot over Network Load Balancing Software networks) that are configured as the primary network interface.
If your system is unable to load the miniroot (or boot over the network), verify that its primary network interface is an Ethernet connection by following these instructions:
Restart the system from the System Maintenance menu. Do not rebuild the operating system during this restart.
Log on and open a shell window.
Use these commands to display the ordering of the network interfaces:
% /usr/etc/netstat -i <primary interface> <secondary interface> ...
If the primary interface is an Ethernet (for example, ef0 or et0), the Ethernet network connection may be dysfunctional. See IRIX Admin: Networking and Mail for information about Ethernet network connections.
If the primary interface is not an Ethernet, go to step 5.
Configure an Ethernet connection as the primary interface.
Reboot the system. When the system is up and running, it should be able to load the miniroot over the network and boot from it.
The snoop(1M) command that captures packets from the network and displays their contents does not support Network Load Balancing Software devices. You have to snoop each and all physical devices in order to snoop a Network Load Balancing Software device.