Skip to main content


Tools Overview

Tools are used to execute steps in a job. Each step indicates what tools should be used to its execution. It can be a shell to execute commands, other programs, or a git tool for cloning Git repository. A step can have several fields with values defined that are required by a tool or are optional.

A tool communicates with Kraken agent using specific protocol. Thanks to this it is possible to write your own custom tools.

There are several ways a tool can get in interaction with Kraken server via Kraken agent. Tools can be:

  • do nothing, just do its job and that's it
  • return list of all available test cases (eg. pytest tool); this allows pre-storing all test case results as not run initially and also splitting execution of test cases to several jobs
  • report continuously test results (eg. pytest tool)
  • report continuously issues (eg. pylint tool)
  • report stored artifacts (eg. artifacts tool)

More about particular tools can be found in Workflow Schema chapter, in Tools section.

Custom Tools

Tools can be written in Python. It is possible to write them in other languages but for Python there is already a framework that makes some actions easier.

Protocol Overview

There is defined a protocol between Kraken agent and a tool. It works differently depending on the direction. Generally an agent invokes one of predefined commands of a tool and then a tool responds to the agent via an HTTP connection using JSON as a response format.

The commands expected by Kraken agent:

  • get_commands - it should return list of commands exposed by a tool; this command must be provided by every tool
  • collect_tests - it returns list of available tests according to specification provided by step definition; optional
  • run - it should just execute a step; optional
  • run_tests - it executes tests according to specification provided by step definition and report results live, progresively, one by one if possible; optional
  • run_analysis - it should perform an analysis and report issues live, progresively, one by one if possible

One of run* commands must be provided by a tool.

Tool Framework

Handling of commands and sending reponses are quite automated by a Python tool framework. The framework required defining one of run* functions and collect_tests function if wanted.

Example of Simple Tool

import os
from . import utils

def run(step, **kwargs):
cmd = step['cmd']
cwd = step.get('cwd', None)
timeout = int(step.get('timeout', 60))
ret = utils.execute(cmd, cwd=cwd, timeout=timeout)
if ret != 0:
return ret, 'cmd exited with non-zero retcode: %s' % ret
return 0, ''

This is a simplified implementation of shell tool. It just executes a cmd command provided in a step. It uses also optional cwd and timeout parameters provided in a step.

Tool framework provides several helper modules. One of them is utils that offers an execute function. To use it impot the module:

from . import utils

A run functions must return an exit code (0 means all is ok) and a message that in case of an error describes the error (otherwise it returns '' empty string).

Example of Testing Tool

import random

def collect_tests(step):
# return list of available tests
tests = ['', 'test.bbb', 'test.ccc']
return tests

def run_tests(step, report_result=None):
# Now the step contains list of tests reported
# by collect_tests function. If there is no collect_tests
# function then the run_tests function should determine
# on its own scope of tests based of step definition.
tests = step['tests']

for test in tests:
cmd = 'command to execute a test'
# status should indicate test result: 0 - not run, 1 - passed, 2 - failed,
# 3 - error, 4 - disabled/skip, 5 - unsupported
result = dict(cmd=cmd, test=test, status=random.choice([0, 1, 2, 3, 4, 5]))
# report_result is a callback provided by Kraken tool framework that sends back
# the test result to the server

return 0, ''

The tool has hardcoded list of tests to execute regardless of step definition. Generally collect_tests function should return list of tests based of step definition. Then the run_tests function receives list of tests in step definition and then it "executes" them one by one, and then reports them using report_result callback from Kraken tool framework.