Scheduler

Purpose

The scheduler is responsible for scheduling the execution of tasks. The execution of those tasks are being prioritized / scored by a ranker. The tasks are then pushed onto a priority queue.

Within the project of KAT, the scheduler is tasked with scheduling boefje, normalizer, and report tasks.

Architecture

See architecture document for the architecture and the extending document for the extending the scheduler with your own custom schedulers, and rankers.

Stack, packages and libraries

Name	Version	Description
Python	^3.10
FastAPI	^0.115.12	Used for api server
SQLAlchemy	^2.0.23
pydantic	^2.11.3
uvicorn	^0.29.0

The scheduler uses PostgreSQL as its database.

External services

The scheduler interfaces with the following services:

Service	Usage
[Octopoes]	Retrieving random OOI’s of organizations
[Katalogus]	Used for referencing available plugins and organizations
[Bytes]	Retrieve last run boefje for organization and OOI
[RabbitMQ]	Used for retrieving scan profile changes, and created raw data in bytes

Project structure

.
├── docs/                           # additional documentation
├── scheduler/                      # scheduler python module
│   ├── clients/                    # external service clients
│   │   ├── amqp/                   # amqp clients
│   │   ├── http/                   # http api clients
│   │   ├── __init__.py
│   │   ├── connector.py
│   │   └── errors.py
│   ├── config/                     # application settings configuration
│   ├── context/                    # shared application context
│   ├── models/                     # internal model definitions
│   ├── schedulers/                 # schedulers
│   │   ├── queue/                  # priority queue implementation
│   │   ├── rankers/                # rankers for tasks
│   │   ├── schedulers/
│   │   │   ├── __init__.py
│   │   │   ├── boefje.py           # boefje scheduler implementation
│   │   │   ├── normalizer.py       # normalizer scheduler implementation
│   │   │   └── report.py           # report scheduler implementation
│   │   ├── __init__.py
│   │   └── scheduler.py            # abstract base class for schedulers
│   ├── storage/                    # data abstraction layer
│   ├── server/                     # http rest api server
│   ├── utils/                      # common utility functions
│   ├── __init__.py
│   ├── __main__.py
│   ├── app.py                      # openkat scheduler app implementation
│   └── version.py                  # version information
└─── tests/                         # test suite

Running / Developing

Typically the scheduler will be run from the overarching nl-kat-coordination project. When you want to run and the scheduler individually you can use the following setup. We are using docker to setup our development environment, but you are free to use whatever you want.

Prerequisites

By the use of environment variables we load in the configuration of the scheduler. See the environment settings section under Installation and Deployment for more information.

Running

# Build and run the scheduler in the background
$ docker compose up --build -d scheduler

Migrations

Creating a migration:

# Run migrations
make revid=0008 m="add_task_schedule" migrations

Sometimes it is helpful to run the migrations in a clean environment:

docker system prune
docker volume prune --force --all

Testing

# Run integration tests
$ make itest

# Run unit tests
$ make utest

# Individually test a file
$ make file=test_file.py utest

# Individually test a function
$ make file=test_file.py function='test_function' utest

Scripts

The scheduler comes with a few scripts that can be used to interact with the scheduler. These scripts are located in the scripts/ directory and can be check its documentation. These scripts are a collection of scripts that are used for various testing and benchmarking purposes.

`load.py`

Allows to create multiple organisations and with a supplied data.csv file create objects on which a select number of boefjes will be performed upon.

docker build -t mula_scripts .
docker run -it --rm --network=host mula_scripts load.py \
    --orgs {number-of-orgs} \
    --oois {number-of-oois} \
    --boefjes {comma-separated-list-of-boefjes}

`benchmark.py`

Allows to benchmark the operations of the Scheduler. When running the load.py the benchmark script can run along side it to measure the performance of the Scheduler.

It will check:

Errors in the logs
Task stats (how many are queued, running, etc.)
CPU and memory usage

docker build -t mula_scripts .
docker run -it --rm --network=host mula_scripts benchmark.py --container {container-id-of-scheduler}