Bladerunner

---

Bladerunner is a program to send and receive information from any type of ssh enabled text based device. It can run through a jumpbox if there are networking restrictions. You can also provide an additional password to use for any program after logging in, such as MySQL or sudo. bladerunner will attempt to use the host password for everything unless you specify otherwise, allowing it to default through sudo in simple use cases. MySQL, FTP, and telnet prompts are included as well as the default Ubuntu and CentOS bash shells and password prompts. You can provide an additional prompt via command line arguments. bladerunner will automatically accept SSH certificates and will throw ^C at any command that exceeds the timeout before returning. Commands can be loaded into a file and run from there line by line per host.

Install

Installation is done via the usual methods:

$ python setup.py build
$ sudo python setup.py install

Alternatively, you can install via pip:

$ pip install bladerunner

Requires

Python (v2.7+), pexpect and futures 2.1.3.

Options

For a full list of options use:

bladerunner --help

Using a file with a list of commands in it is an easy way to execute more complex tasks.

Use of Bladerunner from within Python

It may be useful to run Bladerunner from inside another script. Here's how:

from bladerunner.base import Bladerunner
from bladerunner.formatting import csv_results, pretty_results, stacked_results

def bladerunner_test():
    """A simple test of bladerunner's execution and output formats."""

    # pass in lists of strings, commands and hosts will be executed in order
    servers = ["testserver1.testdomain.com", "testserver2.testdomain.com"]
    commands = ["uptime", "mysql", "show databases;", "exit;", "date"]

    # this is the full options dictionary
    options = {
        "debug": False,
        "delay": None,
        "cmd_timeout": 20,
        "csv_char": ",",
        "extra_prompts": ["core-router1>"],
        "jump_host": "core-router1",
        "jump_password": "cisco",
        "jump_port": 22,
        "jump_user": "admin",
        "output_file": "/home/joebob/Documents/output.txt",
        "passwd_prompts": [],  # usually best to let Bladerunner decide
        "password": "hunter7",
        "password_safety": True,
        "port": 22,
        "progressbar": True,
        "second_password": "super-sekrets",
        "shell_prompts": [],  # this list is typically auto-generated
        "ssh": "ssh",
        "ssh_key": None,
        "stacked": False,  # preference flag for stacked results
        "style": 0,
        "threads": 100,
        "timeout": 20,
        "unix_line_endings": False,
        "username": "joebob",
        "width": 80,  # used in displaying results
        "windows_line_endings": False,  # force the use of \r\n
    }

    # initialize Bladerunner with the options provided
    runner = Bladerunner(options)

    # execution of commands on hosts, may take a while to return
    results = runner.run(commands, servers)

    # Prints CSV results
    csv_results(results)

    # Prints pretty_results using the available styles
    for i in range(4):
        options["style"] = i
        pretty_results(results, options)

    # Prints the results in a flat, vertically stacked way
    stacked_results(results)

Threaded Bladerunner

As of Bladerunner 4.0.0 it is possible to use the run_threaded() method to call the run() method in new thread. This is especially useful inside of Tornado applications, which may need to be responsive in the main thread during a long running task.

It is recommended that you use gen.Task to do this inside of Tornado, but Bladerunner itself simply returns a thread and calls a callback, so it's really up to the implementation as for how the threading is handled. Here's a simple use case for building a non-blocking remote execution function:

from tornado import gen, web
from bladerunner.base import Bladerunner

@gen.engine
def threaded_commands(options, commands, servers, callback=None):
    runner = Bladerunner(options)
    results = yield gen.Task(runner.run_threaded, commands, servers)
    if callback:
        callback(results)

class MyHandler(web.RequestHandler):
    @gen.engine
    def get(self, *args, **kwargs):
        commands = self.qs_dict.get("commands", [])
        servers = self.qs_dict.get("servers", [])
        if commands and servers:
            # password can be a list to try multiple passwords per host
            options = {"username": "root", "password": ["r00t", "d3f4ult"]}
            results = yield gen.Task(threaded_commands, options, commands, servers)
            self.write(200, results)
        else:
            self.write(404, "commands or servers not provided in qs_dict")

Bladerunner Interactive

Sometimes, you need to apply logic to conditionally decide commands to issue based off of the results of a previous command. As of Bladerunner 4.1.0 there are now a couple different ways you can do this.

Single host interactive via python shell

Here is the simplest use case of a BladerunnerInteractive object:

>>> from bladerunner import Bladerunner
>>> runner = Bladerunner()
>>> inter = runner.interactive("some_host")
>>> inter.run("uptime")
'17:46:22 up 23 days, 19:52,  6 users,  load average: 0.17, 0.13, 0.09'

Multiple hosts interactively via python shell

Rather than handling the BladerunnerInteractive objects yourself, you can store them in the base Bladerunner object instead, letting the base object run the interactive command on all hosts in parallel. An example:

>>> from bladerunner import Bladerunner
>>> runner = Bladerunner()
>>> runner.run_interactive("hostname", "some_host")
some_host: some_host
>>> runner.run_interactive("hostname")
some_host: some_host
>>> runner.run_interactive("hostname", "some_other_host")
some_host: some_host
some_other_host: some_other_host

As you can see, supplying more hosts (the second argument, can also be a list), is optional. If you do supply more hosts, they will be added to the internal list. To remove a host from the pool, use Bladerunner.end_interactive() with the hostname or list of hostnames you'd like to remove:

>>> runner.end_interactive("some_host")
>>> runner.interactive_hosts
{'some_other_host': <BladerunnerInteractive object connected to 'some_other_host' at 0xb6f1dd8c>}

Interactive Threading

Both the run and the connect methods of the BladerunnerInteractive objects can be threaded. When using the base object's run_interactive method, it will use multi-threading internally to perform the action on all devices in parallel, but the call itself is blocking. To work around this, you need to use the BladerunnerInteractive objects themselves. An example of threaded connecting and threaded interactive command running:

from tornado import gen
from bladerunner import Bladerunner

options = {}
runner = Bladerunner(options)
inter = runner.interactive("somewhere")
connected = yield gen.Task(inter.connect_threaded)
if connected:
    results = yield gen.Task(inter.run_threaded, "whoami")
    if "root" in results:
        print("god-mode is enabled")
    else:
        print("{} is but a mere plebeian".format(results))
else:
    print("could not connect")

You do not need to make a specific call to connect_threaded, as the run call will detect that it hasn't connected yet and attempt to. However, it may be preferred to know the connection status earlier.

Predefined Interactive Functions

In the instance where you know exactly what you're looking for, and exactly what to do based off of that outcome, it may be easiest to write a BladerunnerInteractive function and let the base object do the threading for you. In this way, we can run the same logic against many hosts. An example script where you need to check the running status of a service and issue a restart on any hosts where the service is currently down:

from bladerunner import Bladerunner

def my_function(session):
    """You can call this anything, but the signature has to be exact.

    You must accept a single non-keyword argument, which will be the
    BladerunnerInteractive object.

    You can return anything you want, anything other than None will be
    returned grouped as a list with all the other function calls.
    """

    results = session.run("/etc/init.d/httpd status")
    if not "is running..." in results:
        session.run("/etc/init.d/httpd restart")
        return session.server

def main():
    runner = Bladerunner({"username": "root"})
    res = runner.run_interactive_function(my_function, ["host1", "host2"])
    print("restarted httpd service on: {}".format(", ".join(res)))

if __name__ == "__main__":
    main()

In the case where you need different connection parameters for multiple sets of devices, make more Bladerunner base objects and spawn the interactive sets off of them. Alternatively, you can call an update on the base object's options, like so:

from bladerunner import Bladerunner

def my_function():
    results = session.run("/etc/init.d/httpd status")
    if not "is running..." in results:
        session.run("/etc/init.d/httpd restart")
        return session.server, True
    return session.server, False

runner = Bladerunner({"username": "user1", "password": "password1"})

# line separated lists of hostnames or IPs can be passed as string filepaths
runner.run_interactive_function(my_function, "/root/server_list_1")

# if you want to end these sessions, remove them from the base object:
runner.end_interactive("/root/server_list_1")

# new BladerunnerInteractive objects inherit the base object's settings
# but you can update them on the base rather than having to make new ones
runner.options.update({"username": "user2", "password": "password2"})

# the connections to these servers will be maintained in the base object
# indefinately! there are also automatic re-connect methods that are used.
# if you need finer grained control of the sessions, you can pool them
# externally to enforce timeouts and/or keepalives.
results = runner.run_interactive_function(my_function, "/root/server_list_2")

# results at this point is whatever we've defined to return in our function,
# inside a list with each function run per host (order not guaranteed).
for server_name, httpd_restarted in results:
    print("httpd on server {} was {}restarted!".format(
        server_name,
        "not " * int(httpd_restarted is False),
    ))

Non-Standard SSH

As of Bladerunner 4.1.8+ you can provide --ssh to give a non-standard command as ssh. This will clear any automatically added flags, so include them in your command if any are required. Also keep in mind if you are also using Bladerunner with a jumpbox, the ssh command needs to be available there as well.

As a usage example for this, Bladerunner 4.1.8+ can be used with the gcloud CLI:

$ bladerunner -nN --ssh="gcloud compute ssh" "echo 'hello world'" $(kubectl get nodes -o name | cut -d '/' -f2 | tr '\n' ' ')

Bugs & TODO

If you come across a bug, please create a new issue in the issue tracking system with enough relevant details and it will be dealt with promptly.

Authoritative Source

Note that this repository is the source repository for the Python Packaging Index and is the upstream repository for all bug fixes and feature development.

This repository is distributed under the GPLv2 license, with the acknowledgement that some (most, for now) of the library source code is under the BSD license and is still Copyright (c) 2015 Activision Publishing, Inc (see the LICENSE file for full details).

Having said that, this project is transitioning away from the prior codebase. To track how much of the code base is GPLv2 vs BSD, you can use:

$ git diff --stat 62d52e04bb86614efc3e6e280b2c9adccddde83f master

It's also being tracked and updated via Travis-CI right here: