The following information describes the 7318 High-Availability (HA) feature that can be used with High-Availability Cluster Multi-Processing (HACMP) servers. HACMP enables the creation of a cluster of loosely coupled systems.
The following provides information on the ICMP version of the 7318 high availability feature that can be used with a HACMP/6000 cluster environment. When there is a redundant network, this feature allows the 7318 to recover from a network failure.
The HA feature is designed to allow a 7318 to physically attach to two distinct local area networks (LANs), using both its 10Base-T and AUI port connectors. Once attached, the 7318 will be given the address of a host or hosts that will determine which network to use.
The 7318 will come online and automatically switch to the interface configured as the primary interface. It will then begin to "ping" the host. As long as those pings are returned, the 7318 will remain on that Ethernet. If, however, the pings stop being returned, the 7318 will switch (after a length of time) to the backup interface. It will then try to ping the host again. If this succeeds, the 7318 will stay on that port. If it fails immediately, or after some amount of time, the 7318 will switch back to primary port and stay there until a ping is returned.
The Internet Control Message Protocol (ICMP) version of the HA uses an ICMP ping packet to determine if the host is online. In the ICMP case, there are separate Internet Protocol (IP) addresses for the host, as well as separate IP and other Internet parameters for both interfaces of the 7318.
Notes:
The following definitions are important in understanding the ICMP version of the HA feature:
The 7318 also has a second set of Internet parameters for the backup interface. These are known as BackupHost, BackupMyIPaddr, BackupMyNetMask and BackupMyDefGateway.
The 7318 will initialize on the primary interface when the HA feature is enabled. Once started, the HA module will ping the primary host IP address using the ICMP protocol, using the ping interval. If the ping is not returned, the switch to backup timer will be started. If that timer expires without receiving any ping response, the HA module will switch the interface to the backup interface.
At this point, all of the Internet parameters for the 7318 will be switched to the backup ones, and a new address (that of the backup host) will be used for pinging. After the pinging starts, if the ping is not returned, the switch to primary timer will be started. If that timer expires without receiving any ping response, the HA module will switch the interface back to the primary interface.
At this point, the HA module will again start pinging the primary host. However, in this state, the 7318 will stay on the primary interface until ping responses are received (for example, the interface will not keep switching in the case where there is no response from either port). Once a ping response is received in this state, the HA module will reset and go back to its initial state where an absence of ping response will again cause the interface to switch.
The [HighAvail] section enables and configures the HA feature for the Model S20.
The [HighAvail] section contains the following entries:
Note: All addresses should be in dotted decimal format.
The ICMP Version of the HA Feature figure graphically shows the layout of a 7318 attached to two separate IP interfaces.
The Model S20 HA configuration uses IP network address format. The IP address consists of four groups of decimal numbers, each group separated by a . (period). Acceptable values for each group are 0 through 255. A typical IP address might look like the following:
255.192.40.16
Configuring the HA feature involves editing the default configuration file and uncommenting appropriate lines.
You must have root user authority.
cp /usr/lib/cns/s20.cfg /usr/lib/cns/7318EtherAddress.cfgwhere the 7318EtherAddress parameter specifies the Ethernet hardware address of the 7318 to be configured for HA.
[HighAvail] enable=0 ;addresstype=icmp ;pinginginterval=60 ;switchtobackup=600 ;switchtoprimary=1200 ;primaryinterface=AUI ;secondaryinterface=10baseT ;primaryhost=0.0.0.0 ;backuphost=0.0.0.0 ;backupmyipaddr=0.0.0.0 ;backupmynetmask=0.0.0.0 ;backupmydefgateway=0.0.0.0
The connection station software model: S20 (3.41.13) is shown in the following figure:
Boot the 7318 using the 10BT with bootp. An example of a .cfg file is shown below:
[HighAvail] enable=1 ;You can select either ICMP (TCP/IP) pings or IPX (NetWare) pings. ;For TCP/IP use address type "icmp". For NetWare, use "ipx". addresstype=icmp ;Timeouts in seconds ; use ratio here pingingInterval:switchToX of at least 5:1 pingingInterfal=5 switchToBackup=30 switchToPrimary=30 ;primaryhost and backuphost are the names of hosts to ping ;to determine when to switch interfaces. The addresses ;can be IP addresses or IPX addresses. In the case of ;IPX addresses, the addresses can be the same. They must ;be different for IP. primaryhost=3.3.3.1 backuphost=2.2.2.2 backupmyipaddr=2.2.2.99 backupmynetmask=255.255.255.0 backupmydefgateway=2.2.2.2 ;The following interfaces should be set to 10baseT or AUI. ;The backupInterface should be the opposite of the primary. primaryInterface=10baseT backupInterface=AUI
There were no special routes in the host, just the routes for the network interfaces that show up when configuring the interfaces with SMIT. From netstat -rn:
default | 129.35.128.1 | UG | 2 | 1132 | tr0 |
2.2.2 | 2.2.2.2 | U | 2 | 48 | en1 |
3.3.3 | 3.3.3.1 | U | 2 | 1349 | en0 |
127 | 127.0.0.1 | U | 1 | 9 | lo0 |
129.35.128 | 129.35.128.61 | U | 7 | 4302 | tr0 |
The route table on the 7318 was interesting before and after the fallover. The show route command showed a loopback route, a default route for the gateway, and a route for the active network interface (3.3.3). After the fallover, all references to the 3.3.3 network were gone and replaced by the equivalent routes for the 2.2.2 network. The original routes were restored with it flipped back over to the primary ethernet.
After the fallover to the secondary ethernet, a telnet session to the second host system at 2.2.2.3 (an AIX Version 4 system) was started. It was lost again when a toggle back to the primary was done.
When the first 7318 in the chain switches interfaces, it switches to the backup ethernet, which is the wrong subnet (IP)/LAN(IPX) for the rest of the 7318s in the chain. The second and subsequent 7318s in the chain cannot switch interfaces to the 10BaseT because it is not connected.
The IPX protocol stack is not set up to use an internal network number for the 7318 (IPX section in the config file). P10-style sessions do not switch over with the telnet sessions, but may recover when the 7318 returns to the primary interface.
There is no dynamic routing in the 7318 to handle the change in routing that would be required.
No option exists in the configuration file to set it or use it. You must rely on the backup gateway to provide a path to the Kerberos server.
No option exists in the configuration to set it or use it. You must rely on the backup gateway to provide a path to the SNMP manager.
Refer to High Availability Cluster Multi-Processing/6000 Administration Guide for more information concerning HACMP.