1、关于 HACMP 在 NIS 环境下使用的注意事项PowerHA SystemMirror HACMP Version 6.1 Administration guide 7x24 maintenance Planning for 7x24 maintenanceNetwork configuration and name servingSetting up and maintaining clear communication paths for the Cluster Manager is a key element for efficient cluster operation.Setti
2、ng up serial networks or other heartbeat pathIt is crucial to have at least one serial network configured for the cluster. Without a serial network, you run the risk of a partitioned cluster if TCP/IP networks fail, since the nodes will be unable to maintain heartbeat communication. You can also use
3、 disk heartbeats or heartbeats over IP aliases to maintain cluster communications.Integrating HACMP with network services（结合网络服务的 HACMP）HACMP requires IP address to name resolution during the configuration process. The three most commonly used methods include: Domain Name Service（DNS） Network Inform
4、ation Service（NIS） Flat file name resolution (/etc/hosts).（Local）By default, a name request will look first for the DNS (/etc/resolv.conf), second for NIS, and last for /etc/hosts to resolve the name. Since DNS and NIS both require certain hosts as designated servers, it is necessary to maintain the
5、 /etc/hosts file in case the DNS or NIS name server is unavailable, and to identify hosts that are not known to the name server. It is required to have all HACMP IP labels in all cluster nodes /etc/hosts tables.To ensure the most rapid name resolution of cluster nodes, change the default order for n
6、ame serving so that /etc/hosts is used first (at least for cluster nodes).To do this, edit the /etc/netsvc.conf file so that this line appears as follows:hosts=local,bindPutting the local option first tells the system to use /etc/hosts first. If your installation uses NIS you can also add nis. For e
7、xample, hosts=local,bind,nisYou can also change the order for name resolution by changing the environment variable NSORDER as follows:NSORDER=local,bind,nis默认的解析顺序为：DNS NIS LocalDNS(/etc/resolv.conf)NISLocal(/etc/hosts)改变解析顺序有两种办法：（1 ） 在/etc/resolv.conf 中加一行：hosts=local,bind,nis（2 ） 修改环境变量 NSORDER 的
8、值：NSORDER=local,bind,nisNote: By default, during the process of IP address swapping, to ensure that the external name service does not cause AIX to map the service IP address to the wrong network interface, HACMP automatically disables NIS or DNS by temporarily setting the AIX environment variable N
9、SORDER=local within the event scripts. If you are using NIS, have the NIS master server outside the cluster, and have the cluster nodes run as NIS slave servers. At a minimum, every HACMP node must be able to access NIS master or slave servers on a local subnet, and not via a router.NIS 主/从服务器需要与 HA
10、CMP 节点同属一个本地子网，不经过路由器。See the Planning Guide and the Installation Guide for information on editing the /etc/hosts file, and also for notes on NIS and cron considerations.Important: You cannot use DHCP to allocate IP addresses to HACMP cluster nodes. Clients may use this method, but cluster nodes can
11、not.Once configuration is complete, name resolution is no longer required and is not used by HACMP cluster services.Tuning networks for best performance调整至最佳网络性能HACMP provides easy control over several tuning parameters that affect the clusters performance. Setting these tuning parameters correctly
12、to ensure throughput and adjusting the HACMP failure detection rate can help avoid “failures“ caused by heavy network traffic.Cluster nodes sometimes experience extreme performance problems, caused by large I/O transfers, excessive error logging, or lack of memory. When this happens, the HACMP daemo
13、ns can be starved for CPU time. Processes running at a priority higher than the RSCT or Cluster Manager subsystems can also cause this problem.The deadman switch is an AIX kernel extension that halts a node when the Cluster Manager does not run for a certain amount of time, usually due to one of the
14、 problems noted above.See Configuring AIX for HACMP in the Installation Guide for information on setting tuning parameters correctly to avoid some of the performance problems noted above./etc/hostsThe cluster event scripts use the /etc/hosts file for name resolution. All cluster node IP interfaces m
15、ust be added to this file on each node.HACMP may modify this file to ensure that all nodes have the necessary information in their /etc/hosts file, for proper HACMP operations.If you delete service IP labels from the cluster configuration using SMIT, we recommend that you also remove them from /etc/
16、hosts. This reduces the possibility of having conflicting entries if the labels are reused with different addresses in a future configuration.Note that DNS and NIS are disabled during HACMP-related name resolution. This is why HACMP IP addresses must be maintained locally.当 HACMP 进行相关名字解析的时候 DNS 和 N
17、IS 将会被禁用。Important: If you manage user accounts with a utility such as Network Information Service (NIS) or Distributed Computing Environment (DCE) Manager, do not use HACMP user management. Using HACMP user management in this environment might cause serious system inconsistencies in the database.如果
18、使用 NIS 或者 DCE 管理用户，不要使用 HACMP 用户管理。For example, an HACMP cluster node may be a client in a Network Information Service (NIS) environment. If this is the case and you need to use the su command to change user id, there must be a route to the NIS master at all times. In the event that a route doesnt e
19、xist and the su is attempted, the application script hangs. You can avoid this by enabling the HACMP cluster node to be an NIS slave. That way, a cluster node has the ability to access its own NIS map files to validate a user ID.Making changes to network interfacesIn some circumstances, you can use
20、the HACMP facility to swap a network service IP address dynamically, to an active boot interface on the same node and network, without shutting down cluster services on the node.自动切换 IPTypically, stop the cluster to make any change to network interfaces. If you must change the IP address of an netwo
21、rk interface, or if you change the IP label/address, make sure to make the changes to both DNS or NIS and the /etc/hosts file. If DNS or NIS and /etc/hosts are not updated, you will be unable to synchronize the cluster nodes or do any DARE operations. If DNS or NIS services are interrupted, the /etc
22、/hosts file is used for name resolution. You must also redo cl_setup kerberos if you are using Kerberos security.If NIS or DNS is in operation, IP lookup defaults to a nameserver system for name and address resolution. However, if the nameserver was accessed through an interface that has failed, the
23、 request does not complete, and eventually times out. This may significantly slow down HACMP event processing. 默认的解析顺序可能降低 HACMP 的性能。To ensure that cluster event completes successfully and quickly, HACMP disables NIS or DNS hostname resolution by setting the following AIX environment variable during
24、 service IP label swapping:NSORDER = localAs a result, the /etc/hosts file of each cluster node must contain all HACMP defined IP labels for all cluster nodes.建议测试：1、 修改环境变量 NORDER 的值；NSORDER=local,bind,nis解析顺序：本地 hosts、DNS、NIS2、 将 NIS 主服务器或从服务器与 HACMP 节点划在同一 vlan 中；3、 直接将 HACMP 节点中的一个作为 NIS 从服务器；4、 直接将 HACMP 所有节点都作为 NIS 从服务器。
