Warn or error on known bad rules configuration #566
Comments
|
I've done some investigation, and it seems like the right answer here is to leverage jsonschema, since YAML and TOML are freely convertible to and from JSON -- at least at the level we use in Refinery config. I also think it's probably best done as a separate tool that someone can run separately, rather than embedding it into the refinery executable. (Although I'd hear arguments to the contrary!) Writing the schema is not technically difficult, but is a bit tedious so is expected to take several days of work. It's now on the list of things to do soon-ish. |
|
I maintain a schema in CUE that could be used for validation as well, if you'd like to see it. It's focused more on code generation than validation, as it's imbibed with the default values used by the Refinery application. |
…#701) Note: this PR is dependent on #700 -- that one should be merged first. ## Which problem is this PR solving? - Adds complete rules validation to the validation system, plus a few bonus features. ## Short description of the changes - Add a rulesMeta.yaml file to describe the rules data. - Extend the validation system to handle a structure nested more deeply than the basic config validation, which was only 2 layers deep. This meant adding objects and arrays of objects. - Add a few more validation features to support the new semantics. - Add a "did you mean...?" feature to suggest the proper spelling if someone gets something wrong. This is particularly useful when you get the case wrong. - Generate a .md file from the rules metadata; it's not yet pretty but it's a good start. Will improve before release. It's been validated against the converted rules_complete rules file. ## Example results: Rule snippet ``` dataset4: RuleBasedSampler: Rules: ``` Validation: ``` Validation Errors in rules file: Within sampler dataset4: unknown group RuleBasedSampler; did you mean RulesBasedSampler? ``` Rule snippet ``` Samplers: dataset1: DynamicSampler: ClearFrequency: 1 SamplerRate: 10 FieldList: - 10 - request.method - http.target - response.status_code UseTraceLength: "true" ``` Validation: ``` Validation Errors in rules file: Within sampler dataset1: field DynamicSampler.UseTraceLength must be a bool Within sampler dataset1: field DynamicSampler.ClearFrequency (1) must be a valid duration like '3m30s' or '100ms' Within sampler dataset1: unknown field DynamicSampler.SamplerRate; did you mean SampleRate? Within sampler dataset1: field DynamicSampler.FieldList must be a string array but contains non-string 10 Within sampler dataset1: the group DynamicSampler is missing its required field SampleRate ``` I'm really happy with this -- it makes it almost impossible to make a major error in your rules. Next up: run the validators on Refinery startup. Closes #566 Closes #559 Closes #518 Closes #273 Closes #266
Similar to #559, but tracking separately...
Today, if you have a typo in rules config like
Sampleinstead ofSampler, refinery will just sample all traces. That can lead to way more event volume than desired until it's fixed.Ideally, if there's a known bad configuration like this, we could either warn or error on it.
The text was updated successfully, but these errors were encountered: