README.rst

connvitals-monitor

Persistently monitors network conditions with respect to a set of specific hosts.

Dependencies

The executable for the connvitals-monitor package (connmonitor) runs on python3 (tested CPython 3.6.3) and requires a python3 interpreter. It also requires `connvitals <https://github.com/comcast/connvitals>`__ to exist as a subdirectory in its directory (or your import path) as it uses that as a library.

Note: Because this package is not in a standard repository (nor is its dependency), dependencies cannot be automatically resolved; you must first install connvitals for this package to work.

Installation

Note: Versions 1.2+ **only* support Linux systems that run systemd. It's possible that it may install on your system even if you do not run systemd, but it will attempt to install the Unit File under /usr/lib/systemd/system.*

Via pip (Standard)

By far the easiest way to install connvitals-monitor is to simply use pip like so:

pip install connmonitor

Note that you'll probably need to run this command with sudo or the --user flag.

Via pip (From This Repository)

The easiest way to install is to simply use pip. You can install directly from this repository without needing to manually download it by running

user@hostname ~ $ pip install git+https://github.com/comcast/connvitals-monitor.git#egg=connmonitor

Note that you may need to run this command as root/with sudo or with --user, depending on your pip installation. Also ensure that pip is installing packages for Python 3.x. Typically, if both Python2 and Python3 exist on a system with pip installed for both, the pip to use for Python3 packages is accessible as pip3.

Manually

To install manually, first download or clone this repository. Then, in the directory you downloaded/cloned it into, run the command

user@hostname ~/connvitals-monitor $ python setup.py install

Note that it's highly likely that you will need to run this command as root/with sudo. Also ensure that the python command points to a valid Python3 interpreter (you can check with python --version). On many systems, it is common for python to point to a Python2 interpreter. If you have both Python3 and Python2 installed, it's common that they be accessible as python3 and python2, respectively.

Finally, if you are choosing this option because you do not have a Python3 pip installation, you may not have setuptools installed. On most 'nix distros, this can be installed without installing pip by running sudo apt-get install python3-setuptools (Debian/Ubuntu), sudo pacman -S python3-setuptools (Arch), sudo yum install python3-setuptools (RedHat/Fedora/CentOS), or brew install python3-setuptools (macOS with brew installed).

Usage

$ connmonitor [input file]
$ connmonitor [ -v --version ]

input file is a file containing a set of options and hosts to check. If not specified, connmonitor will read input of the same format from stdin. If instead -v or --version is passed as the first argument, the program's version is printed to stdout, and the program exits successfully.

Upon receiving SIGHUP (e.g. when the terminal used to run it is closed), connmonitor will attempt to re-read its configuration file, then continue execution. If the configuration file cannot be read, the program will log an error, clean up its resources and exit with error code 1. If input was given on stdin, the program will log an error and resume execution.

connmonitor handles SIGTERM by neatly cleaning up resources (finishing any running tasks and printing their output to stdout, still logging any errors) and exiting.

As a systemd daemon

Starting with version 1.2.1, connvitals-monitor (unfortunately) comes packaged with a systemd Unit File, and will attempt to install it. To run the daemon, simply run systemctl start connmonitor (as root), and to stop it run systemctl stop connmonitor (also as root). By default, the monitor will log its stdout in JSON format to /var/log/connmonitor.log, and its stderr to /var/log/connmonitor.err. Whenever the monitor is started, it looks for a configuration file at /var/run/connmonitor.conf, and creates it if it does not exist with the following default contents (see 'Input Format'):

localhost trace=5000 scan=5000 json=True

The monitor service does not check for filesystem updates to that config file; if you wish to edit it you may safely do so, but should run systemctl reload connmonitor to read in the new configuration.

Input Format

connmonitor expects input formatted like this:

host1 ping=500 trace=30000 hops=40 scan=60000 json=1
host2 ping=50 numpings=10 payload=41
host3 trace=1000 hops=10
...

Note that config lines (except the hostname part, when applicable) are cAsE-iNsEnSiTiVe. Empty lines are ignored at any point.

Each line of the config must begin with a host. This can be either an IP address or a Fully-Qualified Domain Name (FQDN). Currently, IPv6 is not supported, and if an FQDN can only be resolved to an IPv6 address it will not be queried.

After the host, a list of options in the format <name>=<value> can be specified. If an option is not specified, a default value is used. The options and their valid values are:

• ping - can be set to any positive, rational number or 0 (zero). This indicates the frequency at which pings are performed by specifying a duration (in milliseconds) to wait between each burst of pings. A value of 0 indicates that pings should not be sent. Default: 500
• numpings - can be set to any positive integer. Indicates the number of pings that should be sent in a "burst". Default: 10
• payload - can be set to any positive integer. This indicates the size in bytes of a payload to be sent with each ping. Typically, this will have little to no impact on ping results, but can, in some networks/situations diagnose specific issues. Default: 41
• trace - can be set to any positive, rational number or 0 (zero). This indicates the frequency at which route traces are done by specifying a duration (in milliseconds) to wait between each route trace. A new route trace will not begin until the previous one has finished, so setting this to values lower than network latency to the target is typically meaningless. A value of 0 (zero) indicates that route traces should not be performed. Default: 0 (zero)
• hops - can be set to any positive integer. This indicates the number of network hops to be used as an upper limit on route traces. The default value typically suffices in most situations. Default: 30
• scan - can be set to any positive, rational number or 0 (zero). This indicates the frequency at which route traces are done by specifying a duration (in milliseconds) to wait between each port scan. A value of 0 (zero) indicates that port scans should not be done. Default: 0 (zero)
• json - can be set to any integer or 0 (zero), or one of the Python boolean constants: True and False. If this value is any non-zero integer or True, then the output of this host's statistics will be in JSON format rather than the plain-text format. Default: False
• timestamp - can be set to any integer or 0 (zero), or one of the Python boolean constants: True and False. If this value is any non-zero integer or True, then the outputs of this hosts's statistics will always contain timestamps indicating the time at which printing occurs. Default: True

Configuration options can appear in any order and can be separated by any amount/kind of whitespace except for line terminators (Line Feed, Carriage Return, Form Feed etc.). However, the same option cannot be specified multiple times on the same line, even if it always appears with the same value. Furthermore, there must be no space on either side of the = assigning a value to a config variable.

Output Format

connmonitor outputs results to stdout and logs errors to stderr. Output (including JSON output) takes the same form as connvitals, and you can read about that format on that project's README.

Starting with version 3.0, connmonitor will no longer output traces if they are determined to be the same as the most recent route previously recorded for a given host. This is as a result of changes made to connvitals (but only the Python version) which are discussed in greater detail on that project's page.

Starting with version 3.1, connmonitor will output a timestamp as a part of the JSON object (a floating-point number in milliseconds since the UNIX Epoch), and will output a human-readable date and time in the plaintext output on the second line (directly after names/IP addresses) in the system's ctime format. All timestamps are given in the timezone for which the system is configured.

Starting with version 4.0, each statistic is reported individually, and not bundled together the way connvitals outputs them. This essentially looks like a separate output for each statistic, as though each were invoked seperately by a different connvitals invokation. Prior to this version, configurations were global and all statistics were gathered at the same frequency.

Example

Here's an example of a configuration file that will gather port scans and ping statistics for 10 pings per run each having a payload of 1337B - but not route traces - from google.com every half-second with output in connvitals's standard, plain-text output, and do limited port scanning and traceroutes (to a maximum of 30 hops) - but not pings - on the address 127.0.0.1 (localhost) every 50 milliseconds with output in JSON format:

google.com ping=500 payload=1337 scan=500
127.0.0.1 trace=50 json=1 scan=50