Skip to content

RaymiiOrg/bash-http-monitoring

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

31 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

bash-http-monitoring

all is well image

HTTP(s) status dashboard webpage via shell script. Parallel checking, thus very fast, only dependencies are curl, coreutils and bash (version 4 or above). For all of you who want a simple script with a nice webpage to check a few websites. Perfect for a wall mounted monitoring display and a Raspberry Pi.

Installation and configuration is easy to do inside the script. It scales well, both on the checking side as the information display page (dense on purpose). Failed checks appear right on top for you to act on.

You can set an expected status code and a max timeout per check, so if you consider your site up when it returns a 302 (redirect) or 401 (unauthorized) the script consider that okay. If the status code is not what is configured or there is a timeout or another error, the script considers the check failed.

If a check fails, the script will check that specific one again after 5 seconds to prevent flapping.

Author: Remy van Elst

Website with more information

What this does not have:

  • Notifications
  • History

There however is the option to set a callback_url, whenever a check fails the script will send the status to that URL, allowing you to set up your own history logging or alerting.

Changelog

  • 27-12-2020: Initial release
  • 30-12-2020: Added cgi-bin/docker support
  • 11-01-2021: Added callback URL for failed checks

Installation & Configuration

Make sure you have curl installed (apt install curl). If you need a very simple webserver, try micro-httpd, by ACME. (apt install micro-httpd).

Clone the git repository:

git clone https://github.com/RaymiiOrg/bash-http-monitoring.git
cd bash-http-monitoring

Edit the srvmon script and add your sites. A few examples are provided. This is the syntax:

urls[gists]="https://gist.github.com"
urls[lobsters]="https://lobste.rs"
urls[raymii.org]="https://raymii.org"
urls[example]="http://example.org:3000/this/is/a/test"

The first part between the square brackets is the name, the second part between the quotes is the URL you want to monitor. It can be just a domain, an IP or an actual URL, including port and such.

If you want to override the default status code for a check, this is the syntax:

statuscode[gists]=302

The first part between the square brackets must match the urls[] part.

Further global configuration options include:

maxConcurrentCurls=12 # How many curl checks to run at the same time
defaultTimeOut=10 # Max timeout of a check in seconds
flapRetry=5 # After how many seconds should we re-check any failed checks? (To prevent flapping)
title="Status Dashboard" # Title of the webpage 
cgi=false # Enable or disable CGI header
callbackURL="" # leave empty to disable, otherwise see readme

Execute the script and send the output to a file in your webservers documentroot:

bash srvmon > /var/www/index.html

View that file in a web browser.

OS X specifics

On Mac OS X the builtin bash version is too old and the bsd date utility does not support the time in milliseconds option, so you have to use homebrew to install coreutils and bash:

brew install bash coreutils curl

Then you have to run the script with that bash version:

/usr/local/Cellar/bash/#VERSION#/bin/bash

Also in the cronjob if you choose to set that up.

Cronjob setup

If you want to set up a cronjob, send the output to a temp file and when finished, move that temp file over the "actual" file. Otherwise you might end up with an incomplete page when the checks are running. Like so:

* * * * * /bin/bash /opt/srvmon/srvmon > /var/www/index.html.tmp && /bin/mv /var/www/index.html.tmp /var/www/index.html

If the check fails for whatever reason, the "old" page will not be overridden.

CGI header

Some HTTP servers, like Apache, support CGI scripts. To make it brief, these are scripts which are handed a HTTP request to reply to.

The main advantage of using the script as a CGI script, is that the page is generated on demand and as such, provides a live-view on each page load.
If the page is public, this method should be avoided, as it can be easily abused.

If you want to set up CGI mode, you need to copy the script to your server CGI directory. You can use docker to try this out. Like so:

docker run -d -p 9090:80 -v $PWD/srvmon:/usr/local/apache2/cgi-bin/srvmon hypoport/httpd-cgi

Callback URL

This script does not provide other means of alerting or history. If you do want that, you must do a bit of work yourself. The script supports a callback url, whenever a check failed, it will do a POST request to a configurable URL with the status and error. This allows you to setup logging, history, graphs or otherwise alerting yourself. No examples are provided as of yet but feel free to open a merge request.

The JSON sent is in the following format:

{
	"url": "The configured URL, URL encoded",
	"name": "The configured name"
	"expected_status": "The configured expected status code", // as a string
	"actual_status": "The actual status code", // as a string
	"error": "descriptive error text (from curl mostly)"
} 

It also supports slack webhooks. See here on the Slack site how to set up a webhook. The callback URL looks like this:

https://hooks.slack.com/services/T00000000/B00000000/XXXXXXXXXXXXXXXXXXXXXXXX

When the callback URL contains hooks.slack.com, the JSON sent is formatted to match slack's webhook specification.

Each failed check results in its own request. No bundling is done.

You can use HTTPbin to test locally. HTTPbin is a so called echo server, anything that is sent to it is returned for debugging purposes. Set this in the config:

callbackURL="http://127.0.0.1:8888/post/"

Run httpbin in a local docker:

docker run -p 8888:80 kennethreitz/httpbin

Configure some failed checks, either a non matching status code or a non-existing domain:

Example json data:

{
	"url": "https%3A%2F%2Fgist.github.com",
	"name": "gist.github.com",
	"expected_status": "309",
	"actual_status": "302",
	"error": "Status code does not match expected code"
} 

Another example:

{
	"url": "https%3A%2F%2Fwww.buiekjhkjhkhkhkjhnradar.nl",
	"name": "www.buienradar.nl",
	"expected_status": "200",
	"actual_status": "000",
	"error": "curl: (6) Could not resolve host: www.buiekjhkjhkhkhkjhnradar.nl"
}

Screenshots

All checks are okay:

all is well image

A check has failed:

failed checks

Here is how it looks with many hosts (also note how fast it executes, 6 seconds):

many hosts

This is what the early version looked like:

beta

I had this script running at home for at least a year in that form, when I showed it to a friend he liked it, asked me to make it public, but before I did that I polished it up a bit.