HPC Pack ACM diagnostic tools are a set of command line tools for diagnosing HPC clusters. They're distributed in a Python package hpc-acm-cli, based on HPC Pack ACM API.
Python 2.7, 3.5 or 3.6 is required.
There're several ways to install it. Usually, you should use the PyPI. Other ways are mainly for the package development.
This is a standard way to install a Python package.
python -m pip install --user hpc-acm-cli
Note: python may be python2 or python3 for Python 2 or Python 3 for some Linux distributions.
You can install the latest code in development from GitHub by
python -m pip install --user git+https://github.com/Azure/hpcpack-acm-cli.git#egg=hpc-acm-cli
Get the source code to local and then execute
python -m pip install --user -e <path-to-the-source-directory>
Note: the -e option enable the "editable" mode for the package so that any change you do in the source will take effect without reinstallation.
After installation, there're 3 commands avaiable: clusnode, clusdiag and clusrun for checking cluster nodes, checking/doing diagnostic jobs and checking/running general command separately. They each have subcommands, such as list, show new, etc..
If you're using a Python release for Windows, make sure the path of the Scripts direcotry of the Python installation is on the PATH, since the above commands are installed in this direcotry.
Take Python 3.6 for example. By default, it will be installed to C:\Users\<username>\AppData\Local\Programs\Python\Python36. And thus you need to add C:\Users\<username>\AppData\Local\Programs\Python\Python36\Scripts to the PATH.
- Execute a command with
-hparamter to list its subcommands, likeclusnode -h. - For help of a subcommand, say
list, show it likeclusnode list -h. - All these commands require some common parameters. They're
--host,--userand--password. You can save the values for them in a configuration file and thus avoid entering them each time you run a command. See configuration section below for more. - The examples below assume you have the required parameters provided in the configuration file. You could provide them on the command line instead. But if they're missing, you'll encounter an error at runtime.
clusnode is for checking cluster nodes.
For example, to list the nodes in a cluster, execute
clusnode list
By default, it will list 100 nodes at once. If you prefer more, use the --count parameter, like
clusnode list --count 1000
There's also a parameter --last-id for paging. Refer to command help for more.
To check a specific node
clusnode show <node-name>
clusdiag is for checking/doing diagnostic tests on a cluster.
For example, to list available diagnostic tests
clusdiag tests
To run a diagnostic test
clusdiag new <test-name> --pattern <your-node-name-pattern>
The --pattern is a glob pattern just like the file name globbing on most OSes. For example, abc* matches names starting with abc, and thus abc, abc1 and abc2 are all matched. You can use * to match all nodes.
You can also specify several nodes to run the test, by the --nodes parameter, like
clusdiag new <test-name> --nodes "n1 n2 n3"
The nodes named n1, n2 and n3 are specified, spearated by a space and qouted in a pair of ".
To see a list of diagnostic tests
clusdiag list
To check detailed result of a test
clusdiag show <id>
clusrun is for checking/running general command on a cluster.
For example, to run a command on all nodes of the cluster:
clusrun new --pattern "*" "hostname && date"
It will execute hostname && date on all nodes in a cluster.
The above commands share a common configuration file, .hpc_acm_cli_config, for default values for the command line.
The file will be generated at the first time you run any of the commands. It will be put under the user's home directory(~). Typically, it's /home/{username} for Linux, and C:\Users\{username} for Windows.
The configuration file sets default values for command parameters, and the default values can be overriden by those provided on command line. See comments in the file for configurable options and examples.