Skip to content

running scheduled speedtests inside docker and plotting the results with gnuplot

License

Notifications You must be signed in to change notification settings

ansemjo/speedtest-plotter

Repository files navigation

speedtest-plotter

This is a collection of scripts, which takes internet speedtest measurements against the speedtest.net network with taganaka/SpeedTest and plots them with gnuplot. A crontab schedule is used to automate measurements every couple of minutes and save them to a database. The results can be displayed through a simple Flask webserver.

example plot of speedtest results

USAGE

For changes between releases check the changelog.

CONTAINER

GitHub Workflow Status

The main distribution method is the automatically built container ghcr.io/ansemjo/speedtest. Obviously, you need to have a container runtime like docker or podman installed to run the container.

Note: please update your image name to use the Github container registry. I will delete the DockerHub project sometime in the future.

To start the container with default settings run:

docker run -d -p 8000:8000 ghcr.io/ansemjo/speedtest

This will take a measurement every 15 minutes, save them to a SQLite database in /data/speedtests.db and run the webserver on port 8000. Visit http://localhost:8000 to look at the plotted results. (Note: The smoothed bezier curves require at least two measurements and the image will stay blank otherwise. So you might have to wait a while first.)

TIMEZONE

Your local timezone can be set with the TZ environment variable and a string from tzselect. If none is set usually UTC is assumed. For example users in Japan should use:

docker run -d -p 8000:8000 -e TZ=Asia/Tokyo ghcr.io/ansemjo/speedtest

DATABASE

For data persistence, either mount a volume at /data to save the database file or set the environment variable DATABASE to an SQLAlchemy-compatible URI. A PostgreSQL URI might look like this:

docker run -d \
  -p 8000:8000 \
  -e TZ=Europe/Berlin \
  -e DATABASE=postgresql://user:password@hostname:5432/database' \
  ghcr.io/ansemjo/speedtest

SCHEDULE

You can modify the measurement schedule with the environment variables MINUTES and SCHEDULE. The former takes a measurement every n minutes and the latter may define an entirely custom cron schedule like "four times a day":

docker run -d -p 8000:8000 -e SCHEDULE="0 3,9,15,21 * * *" ghcr.io/ansemjo/speedtest

MARKERS AND SCALING

To add horizontal dashed lines in the plot (e.g. to mark your expected bandwidths) you can use environment variables MARKER_DOWNLOAD and MARKER_UPLOAD. The values are given in MBit/s.

In addition or independently from that you can also set a range scaling for the upload plot relative to the download range with UPLOAD_SCALE. For highly asymmetrical connections this makes it easier to see the upload bandwidth. For example, the above example picture was created with:

docker run -d \
  [...] \
  -e MARKER_DOWNLOAD=800 \
  -e MARKER_UPLOAD=40 \
  -e UPLOAD_SCALE=10 \
  ghcr.io/ansemjo/speedtest

DEFAULT FETCH LIMIT

By default, the webserver will fetch the last seven days (7d) for plotting. This can be configured with the limit= query parameter per request and then bookmark this URL; i.e. http://localhost:8000/?limit=30d will fetch the last 30 days. Alternatively, you can set the environment variable FETCH_LIMIT to configure a different default value for all requests without the query parameter above.

FONT AND RESOLUTION

The resolution and font of the SVG output can be configured with environment variables RESOLUTION and FONT respectively. Output resolution is expected as a comma-separated value of x- and y-size; the default is 1280,800. The font can take either only a name (Arial), only a size (,18) or both (Arial, 18). Note that for a font in an SVG to work, the client needs to have the font, not the server. For example:

docker run -d \
  [...] \
  -e RESOLUTION=1920,1080 \
  -e FONT="Fira Sans, 14" \
  ghcr.io/ansemjo/speedtest

SPECIFIC TESTSERVER

If you want to test against a specific server, you can give a host:port combination in the environment variable TESTSERVER. You can use the API at www.speedtest.net/api/js/servers to pick a suitable host key from the JSON; supply a parameter for ?search=... if you need to. By default it lists servers close to you. Note that this is different from the SERVERID used previously! But you can use ?id=... to search for a specific ID.

For example, to test against wilhelm.tel in Norderstedt with the server ID 4087, you'd use:

docker run -d \
  [...] \
  -e TESTSERVER=speedtest.wtnet.de:8080 \
  ghcr.io/ansemjo/speedtest

DISABLE WEBSERVER

The webserver is a single-threaded Flask application and pipes the data to gnuplot in a subprocess, which may not be suitable for production usage. To disable the webserver completely set the PORT environment variable to an empty string. This will only take measurements and save them to the database.

docker run -d -e PORT="" -v speedtests:/data ghcr.io/ansemjo/speedtest

SHORTHAND COMMANDS

To dump the results as CSV from a running container use the dump command:

docker exec $containerid dump > results.csv

To trigger a measurement manually use the measure command:

docker exec $containerid measure

To reimport a previous dump in a fresh container use import:

docker exec -i $containerid import < results.csv

This can also be used to import results obtained manually with speedtest-cli.

PYTHON SCRIPT

You can use the Python script by itself locally, too. First install the requirements:

pip install -r requirements.txt

Choose a database location and take any number of measurements:

./speedtest-plotter -d sqlite:///$PWD/measurements.db measure
...

Then start the flask webserver to look at the results:

TZ=Europe/Berlin ./speedtest-plotter -d sqlite:///$PWD/measurements.db serve

GNUPLOT SCRIPT

To keep things really simple, you can also take measurements manually with speedtest-cli and only plot an image with gnuplot.

The plotscript expects the format that speedtest-cli outputs when using the --csv flag and a header line from --csv-header. To take some measurements manually with a simple sleep-loop:

speedtest-cli --csv-header > results.csv
while true; do speedtest-cli --csv | tee -a results.csv; sleep 600; done
^C

Afterwards plot the results to an SVG picture with:

gnuplot -c plotscript results.csv plot.svg

BREITBANDMESSUNG

If you're in Germany and you have found that your measured speed regularly does not meet minimum contractual obligations ("erhebliche, kontinuierliche oder regelmäßig wiederkehrende Abweichung bei der Geschwindigkeit") and your provider is not responsive to your complaints, you could use the Breitbandmessung App as the next step. It helps you prepare a well-formatted measurement report, which you could use to file a complaint with the Bundesnetzagentur (BNetzA).

LICENSE

Copyright (c) 2019 Anton Semjonov Licensed under the MIT License