When selecting the ping target systems to trigger adapter switching, note the following points:
Adjacency of Network Devices to Server
To maintain redundancy of network devices, set at least two ping targets. If you only set one target, the system becomes unable to communicate if that system suffers a network failure. By setting two or more targets, you can ensure that user communications remain open even when carrying out maintenance on network devices.
Figure 3.2 Adjacency of Network Devices to Server
System with Hot Standby Router
If you set L3 switches (router) of a hot standby configuration as ping targets, set the IP address or virtual IP address for each L3 switch separately.
If you set as a ping target a virtual IP address that is transferred from one L3 switch to another, it is recommended to set the ping interval and number of pings to be greater than the length of time it takes for the virtual IP address to be switched from the failed L3 switch to the standby L3 switch. This is because, if you set a shorter period and one L3 switch fails, the system may assume that both L3 switches have failed.
For more details on ping intervals and number of pings, refer to "6.2.4 hanetpoll."
Figure 3.3 System with Hot Standby Router
Network Devices on Remote Networks
If you set the ping monitor and the monitor target on different networks, the ping monitor status depends on the routing information of the L3 switch (router) linking the server and the monitor target. Accordingly, you should increase the default values for ping monitoring intervals and frequency, taking into consideration the amount of time it takes to disseminate the RIP (Routing Information Protocol) and to update routing information after a network failure. Also, you should add adjacent network devices to the adapter switching parameters as monitor targets, making these settings with care to avoid incorrect switching.
For more details on the settings, refer to "6.2.4 hanetpoll."
Figure 3.4 Network Devices on Remote Networks
Destination Server
If you set a destination server as the ping target, restarting that server for maintenance will be handled as a transmission path error, causing the adapter to be switched. To prevent this, either add a neighboring network device, as well as the destination, to the adapter switch conditions or set so that switching is not carried out even if an error occurs in the destination server.
For more details on the settings, refer to "6.2.4 hanetpoll."
Figure 3.5 Destination Server as Ping Target
After setting ping target, set the detection time of ping monitoring according to the network configuration.
The failure detection time of ping monitoring is equal to the product of the interval and the number of times to monitor. If you set longer failure detection time, the period of stop operation will be longer. If you set shorter failure detection time, packet loss occurs due to the heavily loaded network, and it increases the probability of an error of the detection. Therefore, it is necessary to adjust the ping interval and the number of pings according to the network configuration in advance.
In GLS, when the network is heavily loaded, you can adjust the interval and the number of pings in advance to prevent the detection error of a failure in the transmission path. Generally, the default value will be fine.
See
For more details on the ping interval and the number of pings, refer to "6.2.4.4 param."
You can calculate the GLS' failure detection time (minimum/maximum value) as follows:
Failure detection time (minimum value)
ping interval (second(s)) x (number of pings - 1) + timeout for ping (Note) |
Failure detection time (maximum value)
ping interval (second(s)) x (number of pings - 1) + timeout for ping (Note) + ping interval (second(s)) |
Note) When ping interval is one second, timeout for ping is one second. When ping interval is other than one second, the timeout is two seconds.
The detection time is calculated as follows when setting values are default:
Failure detection time (minimum value): 3 seconds x (5 times - 1) + 2 seconds = 14 seconds
Failure detection time (maximum value): 3 seconds x (5 times - 1) + 2 seconds + 3 seconds = 17 seconds
The following is a sequence of failure in the transmission paths of ping monitoring in NIC switching mode.
Figure 3.6 Sequence of failure in the transmission paths
