Lorikeet

A Parallel test runner for DevOps.

Overview

Lorikeet is a command line tool and a rust library to run tests for smoke testing and integration testing. Lorikeet currently supports bash commands and simple http requests along with system information (ram, cpu).

Test plans are defined within a yaml file and can be templated using tera. Each step within a test plan can have multiple dependencies (run some steps before others) and can have expectations about the output of each command.

Steps are run in parallel by default, using the number of threads that are available to your system. If a step has dependencies either by require or required_by attributes, then it will wait until those steps are finished.

As an example, here's a test plan to check to see whether reddit is up, and then tries to login if it is:

check_reddit:
  http: https://www.reddit.com
  regex: the front page of the internet

login_to_reddit:
  http: 
    url: https://www.reddit.com/api/login/{{user}}
    save_cookies: true
    form:
      user: {{user}}
      passwd: {{pass}}
      api_type: json
  jmespath: length(json.errors)
  matches: 0
  require:
    - check_reddit

( As a side note, we have added jmespath: length(json.errors) & matches: 0 because an invalid login to reddit still returns a status of 200 OK )

And the output of lorikeet:

$ lorikeet -c config.yml test.yml
- name: check_reddit
  pass: true
  output: the front page of the internet
  duration: 1416.591ms

- name: login_to_reddit
  pass: true
  output: 0
  duration: 1089.0276ms

The name comes from the Rainbow Lorikeet, an Australian Bird which is very colourful. Like a canary in a coal mine, lorikeet is meant to provide a way of notifying when things go wrong. Rather than running one test framework (one colour), it is meant to be more full spectrum, hence the choice of a bird with rainbow plumage.

They are also very noisy birds.

Changes in 0.14.0

• Breaking Change: Add in a default timeout (timeout_ms) for http requests to 30000 milliseconds (30 seconds), This default can be changed as per http options below.

Changes in 0.13.1

• Adds slack webhook support. if there are any steps that have errors, this will be sent to a webhook:

lorikeet -s https://hooks.slack.com/services/<your_webhook_here> test

Changes in 0.13.0

• Breaking Change: The run_steps method returns a stream of steps as they complete, rather than waiting for them all to finish
• More Clippy Lints
• Fixed a compilation error on osx

Changes in 0.12.1

• Fixed all clippy issues
• JMESPath strings are no longer quoted
• Added ability to include step output in subsequent requests. You can do this by including ${step_output.<step_name>} where <step_name> is the name of the step to include as output. Currently you will still need to "require" this step

example_output:
  require: another_step
  http:
    url: http://example.com
    body: |
      ${step_output.another_step}

Changes in 0.12.0

• Update to Tokio 1.0
• Updates to all library dependencies

Changes in 0.11.0

• Initial Async Version
• Updates to library dependencies

Changes in 0.10.0

• Upgrade to 2018 crate format
• Fixed terminal painting on Ubuntu 19.10
• A few minor updates to the library version

Changes in 0.9.0

• Upgrade to Reqwest 0.9.x branch, thanks norcali!

• Added multipart support, body, and headers support to the HTTP request type:

To add custom headers, supply a map of header_name: header_value:

Example Header:
  http:
    url: https://example.com
    headers:
      my-custom-header: my-custom-value

Multipart works in the same way as the existing form option, but allows you to also specify files to upload:

Example Multipart:
  http:
    url: https://example.com
    multipart:
      multipart_field: multipart_value
      file_upload:
        file: /path/to/file

You can also just set a generic body via a string:

Example Body:
  http:
    url: https://example.com
    body: |
      This is a generic POST body

Changes in 0.8.0

• The cli app will not panic if there is an issue reading, parsing or running steps, instead it will output a lorikeet step to display what the error is, and still submit it via webhooks, etc..

• Added in initial delay for a step. If you want to wait an arbitrary period of time before running a step, then you can set an initial delay with the delay_ms parameter. This delay is only executed when the step would normally start, so if you have dependent steps, they will run first, then the delay, then the step.

• Added in Retry Policy: If a test fails, you can retry n times by setting the retry_count property. You can also delay retries by setting the retry_delay_ms parameter.

• Both delay_ms and retry_delay_ms are in milliseconds and must be a positive integer value.

• Added initial junit output so you can use lorikeet with jenkins or another CI server that supports junit xml reports. Use -j report.xml to output junit reports.

Changes in 0.7.0

• The main change here was to change the YAML parsing to remove panics, returning a Result<Vec<Step>> which is a breaking change
• A new function get_steps_raw which takes a &str yaml & anything that implements Serialize as a config context. This mainly allows the library to be used without touching the file system for configs or steps. get_steps still can be provided with paths

Installation

Lorikeet is on crates.io, so you can either run:

cargo install lorikeet

Or clone and build this repo:

cargo build --release

Usage

Command line usage is given by lorikeet -h:

USAGE:
    lorikeet [FLAGS] [OPTIONS] [test_plan]

FLAGS:
    -h, --help       Prints help information
    -q, --quiet      Don't output results to console
    -V, --version    Prints version information

OPTIONS:
    -c, --config <config>         Configuration File
    -j, --junit <junit>           Output a JUnit XML Report to this file
    -w, --webhook <webhook>...    Webhook submission URL (multiple values allowed)

ARGS:
    <test_plan>    Test Plan [default: test.yml]

Test Plan

The test plan is the main driver for lorikeet and is already quite flexible. See below for examples and test syntax. By default lorikeet will expect a file test.yml in the current directory.

Config Option

Lorikeet uses tera as a template engine so you can include variables within your yaml test plan. Using -c you can provide the context of the test plan as a seperate yaml file. This file can be in any shape, as long as it's valid yaml.

As an example, say you want to check that a number of servers are up and connected. You can have a config like so:

instances:
  - server1
  - server2
  - server3

And then write your test plan:

{% for instance in instances %}

ping_server_{{instance}}:
  bash: ping -c 1 {{instance}} 2>&1 >/dev/null

{% endfor %}

And run it:

$ lorikeet -c config.yml test.yml
- name: ping_server_server1
  pass: true
  duration: 7.859398ms

- name: ping_server_server2
  pass: true
  duration: 7.95139ms

- name: ping_server_server3
  pass: true
  duration: 7.740785ms

Webhook

You can submit your results to a server using a webhook when the test run is finished. This will POST a json object with the submitter::WebHook shape:

{
    "hostname": "example.hostname",
    "has_errors": true,
    "tests": [{
        "name": "Example Webhook",
        "pass": false,
        "output": "Example Output",
        "error": "Example Error",
        "duration": 7.70
    }]
}

Test Plan syntax

The test plan is a yaml file that is divided up into steps:

<step_name>:
  <step_type>: <options>
  (<description>: <value>)
  (<expect_type>: <value>)
  (<filter_type>: <list or value>)
  (<dependence_type>: <list or value>)

Each step has a unique name and a step type. Optionally, there can be an expect type, and a list of dependencies or dependents.

You can also include a description of what the test does alongside a name, so you can provide a more detailed explanation of what the test is doing

Step Types

There are currently 5 step types that can be configured: bash, http, system, step and value

Bash Step type

The bash step type simply runs the bash command to execute shell scripts:

say_hello:
  bash: echo "hello"

Optionally you can specify not to return the output if you're only interested in the return code of the application:

dont_say_hello:
  bash:
    cmd: echo "hello"
    get_output: false

HTTP Step Type

The HTTP step type can execute HTTP commands to web servers using reqwest. Currently this is a very simple step type but does support status codes and storing cookies per domain.

You can specify just the URL:

check_reddit:
  http: https://www.reddit.com
  matches: the front page of the internet

Or provide the following options:

• url: The URL of the request to submit
• method: The HTTP method to use, such as POST, GET, DELETE. Defaults to GET
• headers: Key/Value pairs for any custom headers on your request
• get_output: Return the output of the request. Defaults to true
• save_cookies: Save any set cookies on this domain. Defaults to false
• status: Check the return status is equal to this value. Defaults to 200
• user: Username for Basic Auth
• pass: Password for Basic Auth
• timeout_ms: Timeout in milliseconds for the request, defaults to 30000 (30 seconds). If set to null or ~ it will never timeout.
• form: Key/Value pairs for a form POST submission. If method is set to GET, then this will set the method to POST
• multipart: Multipart request. Key/Value pairs Like the form option but allows file upload as well.
• body: Like the form/multipart options but a raw string instead of form data for JSON uploads
• verify_ssl: Verify SSL on the remote host. Defaults to true. Warning: Disabling SSL verification will cause Lorikeet to trust any host it communicates with, which can expose you to numerous vulnerabilities. You should only use this as a last resort.

As a more elaborate example:

login_to_reddit:
  http: 
    url: https://www.reddit.com/api/login/{{user}}
    save_cookies: true
    form:
      user: {{user}}
      passwd: {{pass}}
      api_type: json

For Multipart, you can specify files like so:

Example Multipart:
  http:
    url: https://www.example.com
    multipart:
      multipart_field: multipart_value
      file_upload:
        file: /path/to/file

For a JSON upload you can use the body field:

Example Raw JSON:
  http:
    url: https://www.example.com
    body: |
      { "json_key": "json_value" }

System Step Type

The system step type will return information about the system such as available memory or system load using the sys-info crate.

As an example, to check memory:

check_memory:
  description: Checks to see if the available memory is greater than 1gb
  system: mem_available
  greater_than: 1048000

The system type has a fixed list of values that returns various system info:

• load_avg_1m: The load average over 1 minute
• load_avg_5m: The load average over 5 minutes
• load_avg_15m: The load average over 15 minutes
• mem_available: The amount of available memory
• mem_free: The amount of free memory
• mem_total: The amount of total memory
• disk_free: The amount of free disk space
• disk_total: The total amount of disk space

Using the greater_than or less_than expect types means you can set thresholds for environment resources:

system_load:
  description: Checks the System Load over the last 15 minutes is below 80%
  system: load_avg15m
  less_than: 1.6

'Step' Step Type

If you want to make more assertions on the one step, you can use the 'step' step type. This type simply returns the output of the other step:

say_hello:
  value: hello
  
test_step:
  step: say_hello
  matches: hello

This will also implicitly require that the step it gets it output from is run first as a dependency so you don't have to worry about the order.

Value Step Type

The value step type will simply return a value, rather than executing anything.

say_hello:
  value: hello

Filter types

You can filter your output either via regex, jmespath, or remove the output completely. Filters can be provided once off, or as a list, so you can chain filters together:

example_step:
  value: some example
  filters:
    - regex: some (.*)

You can also shorthand provide a filter on the step like so:

example_step:
  value: some example
  regex: some

Note: If the filter can't match against a value, it counts as a test error

Regex Filter

Simply filters out the output of the step based upon the matched value.

say_hello:
  value: hello world!
  regex: (.*) world!

You can either add it as a regex attribute against the step, or in the filter list:

say_hello:
  value: hello world!
  filters:
    - regex: (.*) world!

By default it will match and return the entire regex statement (`hello world!), but if you only want to match a certain group, you can do that too:

say_hello:
  value: hello world!
  regex: 
    matches: (?P<greeting>.*) world!
    group: greeting

This will output simply hello

JMES Path filter

You can use jmespath to filter out JSON documents, returning some or more values:

show_status:
  value: "{\"status\": \"ok\"}"
  jmespath: status

As with regex, this can be part of a filter chain:

show_status:
  value: "{\"status\": \"ok\"}"
  filters:
    - jmespath: status

No Output Filter

If you don't want your output printed in results, you can add no output:

dont_show_hello:
  value: hello
  do_output: false

You can also add this to a filter chain:

dont_show_hello:
  value: hello
  filters:
    - nooutput

Sometimes you might return too much from a request, so you can use this to ensure what's printed out is not included:

check_reddit:
  http: https://www.reddit.com
  filters:
    - regex: the front page of the internet

Expect types

There are 3 expect types currently: Match output, Greater than and Less than. The expect types will take the raw output of the step type and validate against that. In this way you can use it to match against the returned HTML from a web server, or the output of a bash file.

Match Expect type

The match expect type will use regex to match the output of a command.

say_hello_or_goodbye:
  value: hello
  matches: hello|goodbye

If there is an error converting the regex into a valid regex query, then this will be treated as a failure.

Greater than or less than

If your output is numerical, then you can use greater than or less than to compare it:

there_are_four_lights:
  value: 4
  less_than: 5

Dependencies

By default tests are run in parallel and submitted to a thread pool for execution. If a step has a dependency it won't be run until the dependent step has been finished. If there are no dependencies to a step then it will run as soon as a thread is free. If you don't specify any dependencies there is no guaranteed ordering to execution.

Dependencies are important when you need to do things like set cookies before checking API, but will cause your tests to take longer to run while they wait for others to finish.

To defined dependencies you can use the require and required_by arguments to control this dependency tree. The required steps are given by their name, and can either be a single value or a list of names:

step1:
  value: hello

step2:
  value: goodbye
  require: step1

step3:
  value: yes
  require:
    - step1
    - step2

Lorikeet will fail to run and panic if:

• There is a circular dependency
• The step name in a dependency can't be found

Required By

required_by is just the reciprocal of require and can be used where the test plan makes it more readable.

So this step plan:

step1:
  value: hello

step2:
  value: goodbye
  require: step1

Is equivalent to this one:

step1:
  value: hello
  required_by: step2

step2:
  value: goodbye

More complex dependency example

you_say_yes:
  value: yes

i_say_no:
  value: no
  require: you_say_yes

you_say_stop:
  value: stop
  require: 
    - i_say_no
    - you_say_yes
  required_by:
    - and_i_say_go_go_go

and_i_say_go_go_go:
   value: go go go

Retry Counts and Delays

Sometimes you want to delay a step a certain amount of time after another step has been run. Sometimes if a step fails you may also want to retry it a few times before giving up.

Adding a Delay

You can add a delay by setting the delay_ms value:

step1:
  value: hello
  delay_ms: 1000

Output:

$ lorikeet test.yml
- name: step1
  pass: true
  output: hello
  duration: 1004.1231ms

Adding a Retry

You can retry steps a few times with the retry_count and add a delay to the retry by using the retry_delay_ms.

this_will_fail_but_take_3_seconds:
  value: hello
  matches: goodbye
  retry_count: 3
  retry_delay_ms: 1000

Output:

$ lorikeet test.yml
- name: this_will_fail_but_take_3_seconds
  pass: false
  output: hello
  error: Not matched against `goodbye`
  duration: 3015.933ms

JUnit Reports

You can generate a junit xml report with the -j command:

lorikeet -j report.xml test.yml

The output is primarily geared towards using with with Jenkins BlueOcean, and the report format may change a little bit.

Examples

Save these examples as test.yml to run them

Echoing hello from a bash prompt

Test Plan:

say_hello:
  bash: echo hello

Output:

$ lorikeet test.yml
- name: say_hello
  pass: true
  output: |
    hello

  duration: 2.727446ms

Matching the output of a bash command

Test Plan:

say_hello:
  bash: echo hello
  matches: hello

Output:

$ lorikeet test.yml
- name: say_hello
  pass: true
  duration: 2.68431ms

Checking whether reddit is down

Test Plan:

check_reddit:
  http: https://www.reddit.com
  matches: the front page of the internet

Output:

$ lorikeet test.yml
- name: say_hello
  pass: true
  duration: 2.68431ms

Logging into reddit

For configuration parameters of tests such as usernames and passwords, it makes sense to separate this out into a different file:

Config file:

user: myuser
pass: mypass

Test Plan:

login_to_reddit:
  http: 
    url: https://www.reddit.com/api/login/{{user}}
    form:
      user: {{user}}
      passwd: {{pass}}
      api_type: json

Output (Don't forget to specify the config file with -c) :

$ lorikeet -c config.yml test.yml
- name: login_to_reddit
  pass: true
  output: {"json": {"errors": [], "data": {"need_https": true, "modhash": "....", "cookie": "..."}}}
  duration: 1420.8466ms