A linter for Tekton resource definitions
Make sure you've NodeJS v18 or later installed, and run....
npx @ibm/tekton-lint@latest <glob-pattern-to-yaml-files>
To try a quick example, grab the example-task.yaml
from this repo
> wget https://raw.githubusercontent.com/IBM/tekton-lint/main/example-task.yaml
> npx @ibm/tekton-lint@latest example-task.yaml
example-task.yaml
4:9 error Invalid name for Task 'Task-without-params'. Names should be in lowercase, alphanumeric, kebab-case format no-invalid-name
10:14 warning Invalid image: 'busybox'. Specify the image tag instead of using ':latest' no-latest-image
9:31 error Undefined param 'contextDir' at .spec.steps[0].command[2] in 'Task-without-params' no-undefined-param
11:19 error Undefined param 'contextDir' at .spec.steps[0].workingDir in 'Task-without-params' no-undefined-param
✖ 4 problems (3 errors, 1 warning)
You can use the tool as a regular lint tool from the CLI or scripts; alternatively you can use it as a library via it's API or from an IDE such as VScode. (please consider both the API and IDE use as experimental features.)
When I first started using Tekton a few years ago, I found one of the harder things was keeping track of the dependencies between the YAML files. Comparing to say github actions, or travis, Tekton is a lot more verbose. The comparison isn't quite fair as the tools do work at slightly different levels.
Nevertheless, I was pleased to find this TektonLint tool; it was able to spot a number of the 'gotchas' before pushing the pipelines for execution. After a brief hiatus, I was using Tekton again. Bence Dányi the original author was no longer at IBM; as the tool had been useful for me I decided to take on the repo.
I've updated typescript versions, added in additional features such as custom rules, allowed for caching of standard components. Theres more still I'd like to do; but of a MVP it's at v1.
tekton-lint
is parsing the passed files as yaml files, and checks the rules
on the resulting document set. [More details on the pattern syntax.][pattern]
Using tekton-lint
in watch mode will monitor for any changes in the provided paths and automatically run the linter again.
Be mindful if you specify a glob pattern (eg *.yaml
) that this might be expanded by your shell rather than the tool itself. It'll probably be fine for the vast majority of cases.
tekton-lint [<options>] <glob-pattern-to-yaml-files> ...
Options:
--watch Run tekton-lint in watch mode [boolean] [default: false]
--color Forcefully enable/disable colored output
[boolean] [default: true]
--format Format output. Available formatters: vscode | stylish | json
[string] [choices: "vscode", "stylish", "json"] [default: "stylish"]
--quiet Report errors only [boolean] [default: false]
--max-warnings Number of warnings to trigger nonzero exit code
[number] [default: -1]
--config location of the .tektonlintrc.yaml, defaults to cwd
[string] [default: "/home/matthew"]
--refresh-cache If true will delete the cache directory for external tasks
[boolean] [default: "false"]
--version Show version number [boolean]
--help Show help [boolean]
Examples:
tekton-lint "**/*.yaml" Globstar matching
tekton-lint path/to/my/pipeline.yaml Multiple glob patterns
"path/to/my/tasks/*.yaml"
tekton-lint --watch "**/*.yaml" Watch mode
Only the yaml files that are specified are linted; these can be defined either on the command line, or via the .tektonlintrc.yaml
file.
globs:
- example*.yaml
If both command line and the configuration file have patterns specified, both are used but the command line options come first.
Depdending on how your Tekton instance is configured sometimes you might be using task descriptions that aren't part of your project or repository. In this case you'll get linter errors that "Task XYZ can't be found" - the latest v1 version of tool has added the concept of 'external tasks'.
For example if you are using some of the OpenToolchain Tasks, you can add the following to the configuration file
---
external-tasks:
- name: git-clone-repo
uri: https://github.com/open-toolchain/tekton-catalog
path: git
- name: toolchain-extract-value
uri: https://github.com/open-toolchain/tekton-catalog
path: toolchain
- name: doi-publish-build-record
uri: https://github.com/open-toolchain/tekton-catalog
path: devops-insights
- name: icr-publish
uri: https://github.com/open-toolchain/tekton-catalog
path: container-registry
- name: iks-detch
uri: https://github.com/open-toolchain/tekton-catalog
path: kubernetes-service
This will clone the repos specified and extract the various directories to a local cache. When you lint your tools these external tasks and definitions will be included. This will prevent any errors about tasks missing; but also importantly ensure your use of the tasks is correct, eg no missed parameters.
Note that these cached files won't be linted themselves.
The cache is defined to be put to ~/.tektonlint
If you wish to clear the cache, simply delete this directory - alternatively there is a cli option.
Please note that the tool assumes that the files are syntax correct YAML files, and they are following the established schema for Tekton. For example referring to a field with the wrong spelling or case will confuse the rules.
If you see an error like Pipeline 'pipeline-test-perf-tag' references task 'undefined' but the referenced task cannot be found.
referencing 'undefined' this more than likely means there is a syntax error in the yaml file. This can be easily seen with casing, so using Name: ..
rather than name: ..
These rules are straightforward, violations indicate that there's a good chance that you won't be able to run your
Pipeline
- Missing
Task
definitions - Missing
Condition
definitions - Missing
Pipeline
definitions - Missing
TriggerTemplate
definitions - Missing
TriggerBinding
definitions - Missing parameter declarations within
Task
s - Missing parameter declarations within
Pipeline
s - Missing parameter declarations within
TriggerTemplate
s - Missing volume definitions in
Task
s - Missing
Task
outputresult
s - Missing required
Pipeline
parameters inTriggerTemplate
s - Missing required
Task
parameters inPipeline
s - Missing required workspaces of
Task
s referenced inPipeline
s - Missing required workspaces of
Pipeline
s referenced inTriggerTemplate
s - Extra parameters passed to
Pipeline
s - Extra parameters passed to
Task
s - Invalid
runAfter
conditions - Invalid
resourceVersion
key - Duplicate resources (of all supported resource kinds)
- Duplicate parameters name in
Task
s - Duplicate environment variables in
Step
s - Duplicate
PipelineRun
's parameters name inTriggerTemplate
s - Duplicate parameters name in
TriggerBinding
s - Duplicate parameters name in
TriggerTemplate
s - Duplicate parameters name in
Pipeline
s - Missing
Task
parameter value inPipeline
s - Invalid
Task
,Pipeline
,TriggerTemplate
,Condition
parameter names (alpha-numeric characters,-
and_
and can only start with alpha characters and_
) - Invalid
Task
,Pipeline
,TriggerTemplate
parameter value types (must bestring
,multiline string
orarray of strings
) - Invalid
Task
parameter syntax (usingv1beta1
syntax inv1alpha1
defintions, and vice versa) - Invalid (undefined)
Workspace
references inTask
s ofPipeline
s - Missing referenced
Task
in anotherTask
's parameter inPipeline
s - Cycle detection in each pipelines task dependency graph (based on
runAfter
s,results
andresource -> inputs
)
If you violate these rules, the
Pipeline
is probably going to be just fine, these rules are more like a collection of best practices that we collected while we were working with tekton.
- Unused
Task
parameters - Unused
Condition
parameters - Unused
Pipeline
parameters - Unused
TriggerTemplate
parameters - Unpinned images in
Task
steps - kebab-case and OR camelCase naming violations
Task
&Pipeline
definitions withtekton.dev/v1alpha1
apiVersion
- Missing
TriggerBinding
parameter values - Usage of deprecated
Condition
instead ofWhenExpression
- Usage of deprecated resources (resources marked with
tekton.dev/deprecated
label) - Missing
hashbang
line from aStep
sscript
The default rule is for preferring kebab-case; camelCase is equally popular and many of the official examples on the Tekton website are in camel case. To use the rule that accepts camel case as well swap the rules in the .tektonrc.yaml
file
---
rules: # error | warning | off
prefer-kebab-case: off
prefer-camel-kebab-case: warning
You can configure tekton-lint
with a configuration file (.tektonlintrc.yaml
). You can decide which rules are enabled and at what error level. In addition you can specify external tekton tasks defined in a git repository; for example OpenToolChain provides a set of tasks that are helpful. But if you lint just your own tekton files there will be errors about not being able to find git-clone-repo
for example. Not will any checks be made to see if your usage is correct.
Any task defined in the external-tasks
section will be clone from git, into a local cache (defaults to ~/.tektonlint
). Please make sure that you can git clone
before running.
The configuration file should follow this format:
---
rules:
rule-name: error | warning | off
external-tasks:
- name: <tasks name>
uri: http://github.com/....
path: <sub dir of repo>
Example .tektonlintrc.yaml
file:
---
rules:
no-duplicate-param: error
no-unused-param: warning
no-deprecated-resource: off
external-tasks:
- name: git-clone-repo
uri: https://github.com/open-toolchain/tekton-catalog
path: git
- name: toolchain-extract-value
uri: https://github.com/open-toolchain/tekton-catalog
path: toolchain
- name: doi-publish-build-record
uri: https://github.com/open-toolchain/tekton-catalog
path: devops-insights
- name: icr-publish
uri: https://github.com/open-toolchain/tekton-catalog
path: container-registry
- name: iks-detch
uri: https://github.com/open-toolchain/tekton-catalog
path: kubernetes-service
Search path for .tektonlintrc.yaml
- location specified on the command line with the
--config
option - current working directory
- default values used if nothing else found
In additional the default set of rules, custom rules can provided as node modules; the mechanism is similar that used by tools such as eslint.
An example is provided in custom_rules
, and will be used in the examples below. Notes on writing rules are also in the README.md
In the .tektonrc.yaml
file add an object with names of the rules and node module that provides it.
To load rules exported by the module custom_rules
, and named under my_rules
add the custom
field to the yaml file.
---
rules:
....
external:tasks:
...
custom:
my_rules: custom_rules
# For debug and test, refer directly to the js file
# my_rules: ../customer_rules/my_rules.js
A module may report more than one rule. There is one rule in the example, and this flags up any task that starts with the work 'Task' (not meant to be serious rule, but an example!)
Running on the example-task.yaml
file with the example configured adds a 4th report
./example-task.yaml
10:14 warning Invalid image: 'busybox'. Specify the image tag instead of using ':latest' no-latest-image
9:31 error Undefined param 'contextDir' at .spec.steps[0].command[2] in 'Task-without-params' no-undefined-param
11:19 error Undefined param 'contextDir' at .spec.steps[0].workingDir in 'Task-without-params' no-undefined-param
1:1 error Tasks should not start with word 'Task' my_rules#no-tasks-called-task
✖ 4 problems (3 errors, 1 warning)
Within the representation of the rule name here my_rules#no-tasks-called-task
; if you want to turn off this rule, then you can adjust the list of rules in the .tektonrc.yaml
as for another rule; Note that #
in the name.
rules:
...
my_rules#no-tasks-called-task: off
Please raise an issue in the usual manner, please try and include a short sample Tekton file that shows the problem.
There is an internal logging system you can enable via setting and environment variable
TL_DEBUG=true tekton-lint <file.yaml>
This will write out a app.log
file in pino json format - use pino-pretty to view