sense 0.1.2

install

Installs a Python package to the project with pip using the user scheme.

import sense
sense.install(package_name, flags=[], arguments={})

If you prefer, you can also install packages by running a shell command from IPython using the ! prefix:

!pip install pyzmq --user

Options:

• flags: A list of strings to pass to pip as flags. For example, ["U", "use-mirrors"] would translate to the command-line flags -U --use-mirrors.

• arguments: A dict containing arguments to pass to pip. For example, {"d": "./downloads", "mirrors": "http://URL"} would translate to the command-line arguments -d ./downloads --mirrors=http://URL.

Once installed, any of the project’s dashboards can import the package.

network_info

Returns the current dashboard’s contact information in a dict with keys public_dns, public_port_mapping, ssh_password and project_ip.

import sense
network_info = sense.network_info()

Every project has its own virtual private network (VPN). The project IP address is bound to the project VPN and is only accessible to other dashboards in the same project. The project VPN makes it easy to use cluster computing frameworks that lack built-in security features, such as MPI.

The public DNS hostname, public port mapping and SSH password describe how the current dashboard can be contacted from outside the project. The public port mapping is a dict whose keys and values are integers. Only ports that are keys of the public port mapping can be accessed via the public DNS hostname. If you run a service on port 3000, for example, it can be accessed from anywhere on the internet on the public DNS hostname and port public_port_mapping[3000].

If required, you can SSH to dashboards using the public DNS hostname and port public_port_mapping[22] with username “sense” and the SSH password.

launch_workers

Launches worker dashboards into the cluster.

import sense
 worker_info = sense.launch_workers(n,
    size=0,
    engine="sense-ipython-engine",
    script="",
    code="",
    env={})

In Sense, a cluster is a group of dashboards with the same master dashboard. Worker dashboards multiplex their outputs to the master and are cleaned up automatically when the master is stopped or fails. These features make it easy to manage, monitor and debug distributed applications on Sense.

The parameters are:

• n: The number of workers to launch.

• size (optional): The size of the workers, 0 to 16.

• engine (optional): The name of the npm module to use as the engine. Defaults to “sense-ipython-engine”, but workers can run other engines too.

• script (optional): A script file that the worker should execute on launch. The path is relative to the project’s home folder.

• code (optional): Code that the worker should execute on launch. If both are provided, script has precedence over code.

• env (optional): A dict containing environment variables that should be set on the workers before any code is executed. This is the preferred way to send a master’s contact information information to workers.

The return value is a list of dicts. Each dict describes one of the workers that was launched and contains keys such as "id", "engine", "status", etc. The full format is documented here.

list_workers

Returns information on the worker dashboards in the cluster in a list of dicts like those returned by launch_workers.

import sense
worker_info = sense.list_workers()

get_master

Returns information on the cluster’s master dashboard in a dict like the ones returned by launch_workers.

import sense
master_info = sense.get_master()

stop_workers

Stops worker dashboards.

Dashboards’ numerical IDs are available at key id in the dicts returned by list_workers and launch_workers. The return value is a dict of the same type.

import sense

# To stop specific workers:
worker_info = sense.stop_workers(id1, id2, ...)

# To stop all workers in the cluster:
worker_info = sense.stop_workers()

get_auth

Returns authentication information for the REST API as a dict with keys "user" and "password".

Sense’s REST API gives you complete programmatic control over virtually every aspect of Sense. Most REST calls require Basic Authentication. To make authenticated REST calls, supply the information returned by get_auth your HTTP client of choice, such as the Python requests package.

By default get_auth uses the environment variable SENSE_API_TOKEN for authentication. This token restricts access to the current project. For access across projects, you can pass in credentials manually or set SENSE_USERNAME and SENSE_PASSWORD in the environment. To better understand these options, read the Understanding Project Security documentation.

import sense
import requests
auth = sense.get_auth()
url = "https://api.senseplatform.com/users/" + os.environ["SENSE_OWNER_ID"] + \
   "/projects/" + os.environ["SENSE_PROJECT_ID"]
response = requests.get(url, auth=(auth["user"], auth["password"])).json()