ha工作原理（Ha working principle）.docVIP

ha工作原理（Ha working principle） The working principle of HACMP [reprint] The working principle of HACMP The working principle of HACMP is to monitor the host and network card, using LAN state. There is a TCP/IP network and a TCP/IP network in a HACMP environment. TCP/IP network application client to access the public network, the network can be supported by most AIX networks, such as Ethernet, T.R., FDDI, ATM, SOCC, SLIP, etc.. Non TCP/IP network for HACMP on HA (Cluster) in each node to monitor and provide an alternative to TCP/IP communication path, it can be used RS232 serial line to connect each node, each node can also be SCSI card or SSA card set to Target Mode. HACMP will test the fault diagnosis and response to three types: 1 network fault, 2 network, 3 node failure. The following three kinds of faults were introduced. 1, network fault As mentioned above, the cluster structure of HACMP, in addition to the TCP/IP network, and a non TCP/IP network, it is actually a heart, designed to detect node crashes or just network failure. As shown below, once the node joining the Cluster (i.e., the node HACMP has normal start), each card, the node of the non TCP/IP network will continue to receive and send the Keep-Alive signal, the K-A parameters can be adjusted, send a certain number of HA in continuous packets are lost after you can confirm each other card, or network, or node failure. Therefore, with K-A, HACMP can easily find network fault, because once a card failure is sent to the network card K-A will be lost. The node 1 cluster manager (HACMPs brain) will produce a swap-adapter event, and the execution of the event (script HACMP provides the most general environment event scripts, they are using standard AIX commands and HACMP tools to write). Each node has at least two cards, one is service adapter, providing foreign service, the other one is standby adapter, it is only cluster manager and client do not know know, application. Once the swap-adapter event occurs af