#self-hosted #analytics #traffic #low-memory #json-api

app oxitraffic

Self-hosted, simple and privacy respecting website traffic tracker

12 unstable releases (5 breaking)

0.10.3 Aug 24, 2024
0.10.2 May 8, 2024
0.10.1 Jan 28, 2024
0.9.2 Nov 21, 2023
0.3.0 Jul 31, 2023

#113 in Web programming

Download history 1/week @ 2024-07-04 71/week @ 2024-07-25 11/week @ 2024-08-01 167/week @ 2024-08-22 5/week @ 2024-08-29 14/week @ 2024-09-12 55/week @ 2024-09-19 22/week @ 2024-09-26 2/week @ 2024-10-03

81 downloads per month

AGPL-3.0

400KB
2K SLoC

OxiTraffic

Self-hosted, simple and privacy respecting website traffic tracker 🌐

➡️ Demo ⬅️

Features

  • ❌ Short visits are not counted
    • ✅ Only meaningful visits are counted
    • 🤖 Makes it less likely to count visits by web bots
  • 🥷🏼 Respects privacy (no personal data or IP is logged)
  • 🍪 No cookies
  • 🕊️ Self-hosted
  • 📈 Visualization of visits history
  • 💻️ API for visits history and count
  • 🏅 Low memory usage (about 12 MiB)
  • 📦️ First class container support
  • 🔀 Asynchronous and multithreaded
  • 🆓 Free & open source (AGPLv3)
  • 🦀 Written in Rust (oxidized)

Demo

Here is a demo which tracks my own website (mo8it.com).

How it works

You add the following script tag to your website after replacing OXITRAFFIC_BASE_URL with the base URL of your OxiTraffic instance:

<script type="module" src="https://OXITRAFFIC_BASE_URL/count.js"></script>

It runs the tiny script count.js.

The script calls /register?path=PATH to receive a visitor ID. PATH is the path of the page you are on.

This ID is used after the minimum delay (configuration option min_delay_secs) to call /post-sleep/VISITOR_ID which leads to counting that visit.

When the page is left, a request is sent to /page-left/VISITOR_ID to record the total spent time.

Path validation

How does OxiTraffic know if a newly requested path is a valid one for your tracked website?

Only for the first request to a new path, OxiTraffic sends a request to that path prefixed by the configuration option tracked_origin_callback. If the status code is in the range 200-299 (success), the path is added to the database. Otherwise, the request is rejected.

Setup

Containerized

You can use the container image published on Docker Hub.

You can pull that image using Docker:

docker pull mo8it/oxitraffic:latest

Or using Podman:

podman pull docker.io/mo8it/oxitraffic:latest

The container image expects the config file to be mounted as a (read-only) volume at /volumes/config.toml inside the container (a volume doesn't have to be a directory, it can be a file).

By default, the container listens on port 80.

Not containerized

You can also host OxiTraffic directly with the binary that you can install with Cargo:

SQLX_OFFLINE=true cargo install oxitraffic --locked

Make sure to provide the environment variable OXITRAFFIC_CONFIG_FILE when using the binary directly (see the configuration section below).

Database

In both cases (container or binary), you need a PostgreSQL database. There are many guides in the internet that explain how to host one either in a container or directly on the host. You could use my blog post about hosting PostgreSQL using Podman.

Configuration

The binary expects the environment variable OXITRAFFIC_CONFIG_FILE to point to the TOML configuration file config.toml. This environment variable is set to /volumes/config.toml in the container image.

The table below shows the configuration parameters for the configuration file. You can use environment variables to either set or overwrite parameters from the config file.

Parameter Description Default Environment variable
socket_address Use 127.0.0.1:8080 for local testing. 0.0.0.0 is important for usage in a container, but you can pick another port. "0.0.0.0:80" OXITRAFFIC_SOCKET_ADDRESS
base_url The base URL of your OxiTraffic instance. Used to build the count.js script. OXITRAFFIC_BASE_URL
tracked_origin The origin of your tracked website that is used to allow CORS-requests from the count.js script to OxiTraffic. OXITRAFFIC_TRACKED_ORIGIN
tracked_origin_callback The origin of your tracked website that is used to verify a newly requested path as explained above. This option exists to be able to make these requests inside a local network. tracked_origin OXITRAFFIC_TRACKED_ORIGIN_CALLBACK
min_delay_secs Minimum delay in seconds between visiting the website and being able to call /post-sleep to count the visit. It is recommended to call /post-sleep one second after this value. A low value not only counts meaningless visits, but also makes it easier for visits by web bots to be counts. 19 OXITRAFFIC_MIN_DELAY_SECS
db.host PostgreSQL host OXITRAFFIC_DB__HOST
db.port PostgreSQL port OXITRAFFIC_DB__PORT
db.username PostgreSQL username OXITRAFFIC_DB__USERNAME
db.password PostgreSQL password OXITRAFFIC_DB__PASSWORD
db.database PostgreSQL database OXITRAFFIC_DB__DATABASE
utc_offset.hours The hours of your UTC offset 0 OXITRAFFIC_UTC_OFFSET__HOURS
utc_offset.minutes The minutes of your UTC offset 0 OXITRAFFIC_UTC_OFFSET__MINUTES

Example configuration

This is an example of the configuration file config.toml:

# Can be omitted because this is the default value.
socket_address = "0.0.0.0:80"

base_url = "https://oxitraffic.your_domain.com"

tracked_origin = "https://your_domain.com"
# In case both OxiTraffic and your website are in a local network and `website` can be resolved to the local IP address of the your website.
# Omit this option to use the value of `tracked_origin` instead.
tracked_origin_callback = "http://website"

[db]
host = "127.0.0.1"
port = 5432
username = "postgres"
password = "CHANGE_ME"
database = "postgres"

[utc_offset]
hours = 2
# Can be omitted because 0 is the default.
minutes = 0

Logging

Oxitraffic logs to stdout. The default log level is info, but you can choose one of the following other levels:

  • off
  • error
  • warn
  • info
  • debug
  • trace

These levels are sorted from no logging to extensive logging. You can change the default level by setting the environment variable RUST_LOG to the level name.

Endpoints

Dashboard

Endpoint Description Return
/ A list of registered paths to see their visits history. HTML
/stats?path=PATH Statistics of the visits history of a specific path. HTML

JSON API

Endpoint Description Return
/api/counts The visits count for each registered path JSON([{"path": String, "count": i64}])
/api/count?path=PATH The visits count for the specified path JSON(i64)
/api/history?path=PATH The visits datetimes for a specific path with the nullable referrer and global UTC offset. You can use this endpoint to make your own analysis and plots JSON({"utc_offset": String, "visits": [{"registered_at": String, "referrer": Option<String>, "spent_time_seconds": Option<i64>}]})

Script

Endpoint Description Return
/register?path=PATH Register to receive a VISITOR_ID for the PATH (e.g. / or /blog/rust-vs-julia) of the page you are visiting. JSON(u16)
/post-sleep/VISITOR_ID Use the visitor ID after the minimum delay min_delay_secs for the visit to be counted. Only status code 200 on success
/page-left/VISITOR_ID/TIME_ON_PAGE_SEC Use the visitor ID on leaving the page to report the total spent time on the page in seconds (TIME_ON_PAGE_SEC). Only status code 200 on success

Limitations

Number of concurrent visitors

Counting will fail if your website has more than 2^16 = 65536 concurrent visitors.

The cause of this is that the registration ID is assigned periodically. This means that the visitor 65537 will get the ID of visitor 1. When the old visitor tries to communicate with OxiTraffic with that ID, the communication will either fail or will be interpreted as if it was from the new visitor.

This limitation can be avoided, but it would lead to higher RAM usage and slightly worse performance.

That being said, if you really have more than 65536 concurrent visitors, open an issue 😉

Single instance

OxiTraffic is designed as a lightweight single instance app. Although it uses a database, it is still stateful for performance reasons and to make self-hosting easier by avoiding an additional dependency like Redis. A single instance is more than enough for a single website.

The state that prevents having more than one instance for a single website is the cache of visits before reaching the minimum delay.

That being said, if you still need to scale horizontally, open an issue for adding something like Redis 😉

Questions?

Don't hesitate to open an issue ^^

Contributing

You are welcome to contribute to the project!

You can always open an issue. Wait for a response on the issue before starting with a pull request (Rejected pull request are very disappointing).

Use Clippy and rustfmt before submitting code :)

Dependencies

~58MB
~1M SLoC