Machine Stats for Unix-like systems

A simple and effective way to gather machine statistics (RAM, Storage, CPU) from a server environment as a first layer of a Tidal discovery process.

Machine Stats for Linux/Unix leverages Ansible to gather facts in a cross-platform way.

Interactive tutorial

Get familiar with Machine Stats, Tidal Tools and Tidal Accelerator!

Installation

Install locally in a Python 3 environment:

python3 -m pip install machine-stats

Need to install in an environment without internet access? Checkout how to do that below.

Data captured

For Linux/Unix based systems, by default, the following metrics are captured from the resources and sent and stored in Tidal Accelerator:

• Host Name
• FQDN
• IP Addresses
• RAM Allocated (GB)
• RAM Used (GB)
• Storage Allocated (GB)
• Storage Used (GB)
• CPU Count
• Operating System
• Operating System Version
• CPU name

Here is an example of the output of machine-stats:

{
    "servers": [
        {
            "cpu_count": 2,
            "cpu_name": "Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz",
            "custom_fields": {
                "cpu_average": 0.9314873092946548,
                "cpu_peak": 20.46949490073019,
                "cpu_sampling_timeout": 30,
                "cpu_utilization_timestamp": "2024-05-17 11:54:19"
            },
            "fqdn": "ip-172-31-9-62.ca-central-1.compute.internal",
            "host_name": "ip-172-31-9-62",
            "ip_addresses": [
                "172.31.9.62",
                "172.17.0.1",
                "fe80::4ee:9cff:fe97:418f"
            ],
            "operating_system": "Ubuntu",
            "operating_system_version": "18.04",
            "ram_allocated_gb": 0.9267578125,
            "ram_used_gb": 0.7001953125,
            "storage_allocated_gb": 7.80632209777832,
            "storage_used_gb": 2.61613130569458
        }
    ]
}

It's also possible to capture point-in-time CPU utilization using the flags --cpu-utilization-only-value and --cpu-utilization-timeout.

Here's an example of the output of running machine-stats hosts --cpu-utilization-only-value --cpu-utilization-timeout 1:

{
    "servers": [
        {
            "cpu_count": 2,
            "cpu_name": "Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz",
            "custom_fields": {
                "cpu_sampling_timeout": 1,
                "cpu_utilization": 0.497512437810943,
                "cpu_utilization_timestamp": "2024-05-17 11:59:54"
            },
            "fqdn": "ip-172-31-9-62.ca-central-1.compute.internal",
            "host_name": "ip-172-31-9-62",
            "ip_addresses": [
                "172.31.9.62",
                "172.17.0.1",
                "fe80::4ee:9cff:fe97:418f"
            ],
            "operating_system": "Ubuntu",
            "operating_system_version": "18.04",
            "ram_allocated_gb": 0.9267578125,
            "ram_used_gb": 0.7109375,
            "storage_allocated_gb": 7.80632209777832,
            "storage_used_gb": 2.6161465644836426
        }
    ]
}

You also can optionally capture metrics about processes running on the server:

• User
• Process Name
• Process Path
• Memory Used (MB)
• Max Memory Used (MB)
• Total Alive Time

To enable capturing process metrics add the command-line flag --process-stats:

machine-stats --process-stats

Here is an example of the output of this command:

{
    "servers": [
        {
            "cpu_count": 2,
            "cpu_name": "Intel(R) Xeon(R) Platinum 8259CL CPU @ 2.50GHz",
            "custom_fields": {
                "cpu_average": 0.4973174560230535,
                "cpu_peak": 13.43191223297967,
                "cpu_sampling_timeout": 30,
                "cpu_utilization_timestamp": "2024-05-17 11:56:39"
            },
            "fqdn": "ip-172-31-9-62.ca-central-1.compute.internal",
            "host_name": "ip-172-31-9-62",
            "ip_addresses": [
                "172.31.9.62",
                "172.17.0.1",
                "fe80::4ee:9cff:fe97:418f"
            ],
            "operating_system": "Ubuntu",
            "operating_system_version": "18.04",
            "process_stats": [
                {
                    "max_memory_used_mb": 283.7890625,
                    "memory_used_mb": 220.11328125,
                    "name": "systemd",
                    "path": "/",
                    "pid": 1,
                    "ppid": 0,
                    "total_alive_time": 420,
                    "user": "root"
                }
            ]
        }
    ]
}

Minimal example

1. Create a hosts file in the current directory. See below on a couple ways you can easily create this.

2. Add connection strings in the form of ssh-user@ip-address or ssh-user@domain to the hosts file one per line If the ssh-user@ part is omitted, then the current user name is used.

3. If you need to use a custom SSH identity file for some particular host, provide it as the following:

   my-user@example.com ansible_ssh_private_key_file=path/to/key-file.pem
   

4. Make sure that Python 2.6+ is installed on the machines from hosts file.

5. If python executable was installed into non-default location (not in /usr/bin/python), add the ansible_python_interpreter parameter to the hosts file after the host IP/domain, for example:

   freebsd.example.com ansible_python_interpreter=/usr/local/bin/python
   

6. Execute machine-stats and pipe its output to Tidal Tools:

   machine-stats | tidal sync servers
   

Additional notes

By default Machine Stats looks for the hosts file in current working directory. If your inventory file has another name or is located on another path, then you should specify it explicitly:

machine-stats /path/to/myhosts | tidal sync servers

You can specify multiple inventory files as the following:

machine-stats hosts myhosts /path/to/myhosts

Configuration

Machine Stats uses Ansible under the hood. Most of the Ansible configuration options can be used with Machine Stats too. By default, Machine Stats will look for configuration files in the following locations:

• $PWD/machine_stats.cfg
• $PWD/machine-stats.cfg
• $PWD/machinestats.cfg
• $PWD/ansible.cfg
• $HOME/.machine_stats.cfg
• $HOME/.machine-stats.cfg
• $HOME/.machinestats.cfg
• $HOME/.ansible.cfg
• /etc/ansible/ansible.cfg

Also, it is possible to specify the custom configuration file location by setting the ANSIBLE_CONFIG environment variable, for example:

ANSIBLE_CONFIG=/path/to/my/machine_stats.cfg machine_stats /path/to/my/hosts

Note: if ANSIBLE_CONFIG value points to a directory, then Machine Stats will look for ansible.cfg in that directory.

Offline installation

NOTE: Creating the packages archive for offline installation and the actual offline installation process must be performed on machines with the same OS and Python versions.

1. On the machine with internet connection create the packages archive using the following commands:

   python3 -m pip download -d machine-stats-offline machine-stats
   tar czf machine-stats-offline.tar.gz machine-stats-offline
   

2. Transfer the archive to the machine where you need to perform the offline installation (replace <remote-host> and <remote-dir> with the appropriate values):

   scp machine-stats-offline.tar.gz <remote-host>:/<remote-dir>/
   

3. On the remote host, extract the archive and switch to extracted directory:

   tar xf machine-stats-offline.tar.gz
   cd machine-stats-offline
   

4. Install Machine Stats and its dependencies:

   python3 -m pip install --no-index --find-links . machine_stats-*.whl
   

Generating a hosts file from Tidal Accelerator

You can easily generate a hosts file directly from your server inventory in Tidal Accelerator. For example you can use this command:

tidal export servers | jq '.[].host_name' > hosts

This will create a file (hosts), in your current directory, that you can use above in Step 1.

Alternatively, if you use Tidal Accelerator Ansible Tower integration script you can use its output to generate the hosts file for machine_stats.

Requirements

• jq

Usage

cd ansible-tower-integration
./tidal_inventory.py | jq -r '.servers.hosts[]' > path/to/hosts

Troubleshooting

machine-stats: command not found

If running Machine Stats as a CLI failed, try running it as the following:

python3 -m machine_stats