qlog-dancer parses logs and generates stats and graphs.

Supports log files in the following formats:

• qlog and sqlog format according to version 0.3 of the datamodel.
• Chrome netlog format

On Debian-based systems, the following additional packages are required:

• cmake
• libexpat1-dev
• libfreetype6-dev
• libfontconfig1-dev

qlog-dancer can chew through logs and produce reports in text or HTML.

The following example creates a directory named "chrome-net-export-log.json-report" and places HTML files containing reports there.

$ cargo run --release -- --report-html /path/to/chrome-net-export-log.json

The following example prints out text reports to stdout

$ cargo run --release -- --report-text /path/to/chrome-net-export-log.json

Both the --report-text and --report-html options will produce a table summary of the qlog events contained in the qlog file. This table can be filtered using the --qlog-wirefilter option, which uses Wirefilter for Wireshark-like expressions and matching behavior. The fields supported are:

• category: A string representing the qlog category of the event
• name: A string representing the qlog event name without category.
• stream_id: An integer representing a QUIC stream ID, that can appear in many types of event. A filter based on stream ID will match several events.

Some examples:

• --qlog-wirefilter 'name != "data_moved" && name != "metrics_updated"' will filter out events with the name data_moved and metrics_updated
• --qlog-wirefilter 'category = "http"' will filter in events belonging to the http category.
• --qlog-wirefilter 'any(stream_id[*]==0)' will filter in events that contains a stream ID of 0.
• --qlog-wirefilter 'any(stream_id[*] in {0 3})' will filter in events that contains a stream ID of either 0 or 3.

When using the --report-html option, a table of filtered qlog events in the provided file will be produced in event-list.html. This includes an in-browser search capability based on the default DataTables library.

qlog-dancer can chew through logs and produce charts as PNG images.

By default, this feature is not enabled. Use the charts option to enable it, use -h for a description of the options.

The following example creates a directory named "file.sqlog-charts" and places images there.

$ cargo run --release -- --charts all /path/to/file.sqlog

There are several types of chart including:

• file.sqlog-conn-overview.png
  • Congestion control, RTT, and combined stream data plotting
• file.sqlog-conn-spark-absolute.png
  • Plot individual charts for stream data buffering and STREAM frame emission, x-axis (representing time) intercept is test run start time
• file.sqlog-conn-spark-relative.png
  • Plot individual charts for stream data buffering and STREAM frame emission, x-axis (representing time) intercept is when stream buffering started

qlog-dancer uses the plotters library, which has interpolated line series. We attempt to work around this in a pre-processing step but be careful to check line series rending matches the expectations of underlying daa types.

Netlogs can contain events related to many connections. The netlog-filter option is a comma-seperated list of hostnames to filter in to netlog analysis. By default, all hostname are analyzed.

$ cargo run --release -- --netlog-filter "example.com" /path/to/chrome-net-export-log.json

qlog-dancer also provides some capabilities as a web app via WASM. Some of the file handling and loading code is different but otherwise it still uses plotters to generate charts that are then drawn to a canvas on a web page. This provides support for interactive elements with the plots.

For local development:

1. Install wasm-pack.
2. From the qlog-dancer directory, run wasm-pack build --target=web. This generates files into a pkg subdirectory.
3. Launch a webserver of any kind that can serve the checked in index.html and the generated file such pkg/qlog_dancer.js. For example, the following works just fine

path/to/qlog-dancer/qlog-dancer$ python3 -m http.server
Serving HTTP on 0.0.0.0 port 8000 (http://0.0.0.0:8000/)

4. Connect to your webserver and click the "Choose file" button to load a file and render some plots. E.g. http://localhost:8000/ (no need to type index.html since the example server deals with that itself)