|
| |
There are two ways to add a node into cluster. One is after SIC installation, you can answer "yes" and start the setup program. Another way is running shell command "slicsetup", located at "/opt/slic/bin".
1. Login to the system as root user
2. Execute command "/opt/slic/bin/slicsetup" to setup a node
3. Welcome message shows and you can either create a new cluster or join a cluster.
Input "c" then press [ENTER] in order to create a new cluster. If you need to join an existing cluster, please read section - adding node to an existing cluster for more detail.
node1:~ # /opt/slic/bin/slicsetup ShaoLin InfiniCluster v2.0-2rc4 setup tools Copyright (c) 2004 ShaoLin Microsystems Ltd. All rights reserved. +++++++++++++++++++++++++++++++ Setup +++++++++++++++++++++++++++++++ Warning: any change of IP Address would be affected the configuration of the cluster, please verify your information before input. +++++++++++++++++++++++ Create or Join Cluster +++++++++++++++++++++++ If you want to create a new cluster, please press 'c'. If you have already created a cluster and want to add current node to the cluster, please press 'j' to join current node to the existing cluster. Create a new (c)luster or (j)oin an existing cluster:c |
4. You need to input a cluster key for the cluster. Cluster key is used to verify the communication between nodes within the cluster. Each cluster key contains up to 15 characters. Each cluster with unique cluster key is highly recommended. Cluster key can be up to 63 characters, including space.
++++++++++++++++++++++++++++ New Cluster ++++++++++++++++++++++++++++ Setting up new cluster... ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ You are required to define a cluster key for your cluster. ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ The purpose of cluster key is similar to a password for each node to authenticate each other, such that fake node cannot break into the cluster. Please input cluster Key: cluster1_key1 |
5. Each node uses a node id to represent in the cluster. Node id contains up to 63 characters. Each node id should have an unique node id.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ setup a new node... ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ The 'Node ID' of a cluster node is for uniquely identify a node in the cluster. Therefore, please make sure the 'Node ID' is not duplicated with other member. Please input Node ID: cluster1_node1 |
6. Select the network interface with the IP address that represent the public network for cluster initiative, this IP address is used to determine the network connection between nodes.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ The 'Public IP Address' is for other node to communicate with current node in some special situation. Therefore, please make sure other nodes can access the current node via this IP address. ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Please select the public IP address: 1) eth0 = 192.168.3.131 2) eth1 = 192.168.111.131 3) eth2 = 192.168.112.131 Please input choice: 1 |
7. Enter local node system reset mode
Select local node system reset mode********************************************************************** Please select system reset mode for the local node. The system reset mode is used to immediate stop this node in failure situations List of available reset modes, [0] none (disabled, not recommended) [1] restart [2] halt [3] poweroff [4] panic Please make your selection [4]: Selected reset mode is "kernel panic", are you sure?(y/n)[Y]: |
8. Setup Clusterwide communication channel.
********************************************************************** Setting up clusterwide communication channel... Please select a cluster type, [1] Ethernet (This uses Ethernet as communication channel) [2] IP (This uses IP as communication channel) Please input selection 1 |
9. You need to create a cluster wide communication channel in order to setup heartbeat between nodes within the cluster. You can select from Ethernet or IP as the communication channel interface.
When you selected '1' Ethernet communication channel, following task will need to perform;
1. You need to input a "Keep alive time" for this communication channel. The default keep alive time is 3000ms.
2. You need to input a"die timeout" for this communication channel. The default die timeout is 9000 in ms.
++++++++++++++++++ Keep alive and die timeout setup ++++++++++++++++++ The 'keep alive time' is the period of sending heartbeat out repeatedly. Keep alive time (in ms, Default = 3000): 3000 The 'die timeout' is the timeout parameter for determinate the heartbeat failure. If a node fail to received the heartbeat from other node X longer than the 'die timeout', then node X on this communication channel is considered as failure. Die timeout (in ms, Default = 9000): 9000 |
3. You can setup this channel so that when a node failure on this communication channel, it will be consider as node failure immediately.
4. You can select the available network interface as Ethernet communication channel for the cluster.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ If node X is fail on a communication channel, node X will be considered as fail immediately. ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ set this communication channel?(y/n) y Please select a network interface for Ethernet communication channel: 1) eth0 = 192.168.3.131 2) eth1 = 192.168.111.131 3) eth2 = 192.168.112.131 Please input choice: 2 |
10. SIC requires at least 2 communication channels in order to start the cluster, you need to give an ID for second communication channels.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Setting up SECONDARY cluster communication channel... ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Please input communication channel ID: cluster1_ccid1 |
You can select IP as communication channel by selecting "2".
1. You need to input a "Keep alive time" for this communication channel, The default keep alive time is 3000 ms.
2. You need to input a"die timeout" for this communication channel. The default die timeout is 9000 ms.
Please select a cluster type, [1] Ethernet (This uses Ethernet as communication channel) [2] IP (This uses IP as communication channel) [3] TTY (This uses TTY as communication channel) Please input choice 2 ++++++++++++++++++ Keep alive and die timeout setup ++++++++++++++++++ The 'keep alive time' is the period of sending heartbeat out repeatedly. Keep alive time (in ms, Default = 3000): 3000 The 'die timeout' is the timeout parameter for determinate the heartbeat failure. If a node fail to received the heartbeat from other node X longer than the 'die timeout', then node X on this communication channel is considered as failure. Die timeout (in ms, Default = 9000): 10000 |
3. You can setup this channel so that when a node failure on this communication channel, it will be consider as failure immediately.
4. You can select the available network interface as Ethernet communication channel for the cluster.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ If node X is fail on a communication channel, node X will be considered as fail immediately. ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ set this communication channel ?(y/n)n Please select the IP address for IP communication channel: 1) eth0 = 192.168.3.131 2) eth2 = 192.168.112.131 Please input choice: 2 |
11. If you need to create more communication channel, you can answer "yes" to create additional communication channel for this cluster. Answer "no" will finish the communication channel setup.
11. A setup success message shows up and you can start the cluster manager with answering "yes" or you can answer "no" to start it later by using "/etc/init.d/slic/slicmgr start" command.
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Create more communication channel?(y/N)n Setup success! Do you want to start the cluster manager now?(y/n)y |
NOTE: if you do not have a valid license, a license verification error will be present.
Starting ShaoLin InfiniCluster Manager: slicmgr: Can't verify license. |
Remember, you need to do installation and node setup on each system that you plan as a cluster member.