Skip to content

Latest commit

 

History

History
77 lines (49 loc) · 2.2 KB

README.md

File metadata and controls

77 lines (49 loc) · 2.2 KB
Raw

httplog - http log processor

httplog is a CLI for processing http log files and providing metrics and alerts.

Requirements:

  • Python 3.8

Running

To run locally in development:

python3 -m venv my_env
source my_env/bin/activate

make test
make dev

httplog -h

To run in a docker environment:

make build

docker run --rm -v <path_to_file_locally>:<path_to_file_in_container> httplog:latest httplog --input-file=<path_to_file_in_container>

To install as a package:

make test
make install

httplog -h

Design

httplog is designed as an extendable pipeline.

The Pipeline class is a wrapper class that expects a list of callables to process. These callables extend the Stage class and must implement it's methods.

The expectation for the pipeline are that:

    - First stage must consume string
    - Last stage is a sink and not a stage
    - Stages publish types that can be consumed by next stage

This allows for composition of the sort:

Pipeline([
        lambda sink: HttpParser(sink),
        lambda sink: Metrics(sink),
        lambda sink: Alert(threshold, timedelta(minutes=2), sink),
        print
])

Implementation

Three separate stages were implemented.

    - Stage 1 (HttpParser): This stage takes a line as input and parses it in accordance with the expected http log format and forwards it to the next stage.
    - Stage 2 (Metrics): This stage updates the current snapshot view of the metrics with the new information and forwards a copy to the next stage.
    - Stage 3 (Alerts): This stage keeps a view of the latest alert window and returns a string on threshold trigger or recovery.
    - Stage 4 (Sink): This final stage is just a sink, which in this case prints the alerts.

Finally, the Pipeline class also launches a thread to periodically log the state of the metrics. The Metrics stage stores a new snapshot in a "mocked" in-memory database, so it can be fetched by this thread.

In order to facilitate usage a module named httplogcli was also implemented, to provide a CLI for interacting with the pipeline. The implementation of the actual pipeline logic is in a separate module, so it can be re-used.