Parallel Shell
The parallel shell feature in Bright Cluster Manager™ allows for simultaneous execution of commands or scripts on multiple nodes in the cluster.
The command or script can be run on all slave nodes in the cluster or on all slave nodes in a
node groupA node group is a user-defined group of nodes..
How It Works
The parallel shell allows for simultaneous execution of commands or scripts on multiple nodes in the cluster.
When a command is run using the parallel shell, the following happens:
- The command is sent to all cluster management daemons (CMDaemons) on the selected slave nodes, in parallel.
- The CMDaemons execute the commands and collect the output, in parallel.
- The CMDaemons return the output to the CMDaemon on the head node, in parallel.
- The output is displayed, either grouped per node, or in the order it arrives.
You can have multiple parallel shells running commands in parallel.
The parallel shell has some advanced features that give you insight in and control over the commands running on the nodes.
The parallel shell has some advanced features that give you insight in and control over the commands running on the nodes.
The screenshot close to this text illustrates this. It shows the PIDs (Process IDs) and the number of CPU cycles associated with the commands.
Using the Parallel Command Details dialog, it is also possible to kill the selected commands.
The Advantage of the GUI
Being able to see the output grouped per node — as it arrives — is very convenient, but only possible in the cluster management GUI. With any command-line based parallel shell, you have to either wait for all commands to complete before the results can be displayed, or the output is displayed as it arrives, but in random order.
The GUI also allows you to see which commands are currently being run by the parallel shell.
|