Skip to content

Latest commit

 

History

History
145 lines (103 loc) · 3.34 KB

README.md

File metadata and controls

145 lines (103 loc) · 3.34 KB

tsqtsq - A PromQL Query Library

tsqtsq aims to make hard-coded PromQL queries easier to read and maintain. Wide-ranging changes and common "query snippets" have varying approaches and often impact query readability. By introducing typed, templated queries, changing common expressions and debugging becomes much easier.

Consider the following use cases:

  • Implement de-duplication of all existing queries
  • Establish patterns for writing new queries
  • Create reusable snippets that accept arguments for metric labels

The library in this directory is an effort to reduce the potential toil involved in refactoring tasks like those mentioned above.

Principles

  • Maintain "backwards compatibilty" by returning PromQL queries as a simple string - just like the string literals and template strings we used before.
  • Re-usability of "query snippets" is a priority.
  • Avoid verbose library usage and syntax wherever possible, prefer ease of use over type purity.
  • Aim to make metrics and labels "discoverable" through IntelliSense, to aid query writing in the editor.
  • Embed "sensible defaults" and tribal knowledge using query abstraction - e.g. using container!="" as a default matcher for requests/limits but only if the container label is not passed a value (this avoids matching against the confusing pod-level cgroup metrics).
  • Prefer named object/property parameters over ordered/implicit arguments - because who can remember whether the labels or the query comes first.

Examples

import { promql, Expression } from 'tsqtsq';

sum

promql.sum({ expr: 'test_metric{foo="bar"}', by: ['foo', 'bar', 'baz'] });

becomes

sum by (foo, bar, baz) (test_metric{foo="bar"})

rate

promql.rate({ expr: 'test_metric{bar="baz"}', interval: '5m' });

becomes

test_metric(foo{bar="baz"}[5m])

label manipulation

promql.label_replace({ expr: 'test_metric{foo="bar"}', newLabel: 'baz', existingLabel: 'foo' });

becomes

label_replace(test_metric{foo="bar"}, "baz", "$1", "foo", "(.*)")

aggregation over time

promql.sum_over_time({ expr: 'test_metric{foo="bar"}' });

becomes

sum_over_time((test_metric{foo="bar"})[$__range:])

simple offset

promql.offset({ units: { d: 42 } });

becomes

offset 42d

complex offset

promql.offset({ units: { y: 2, d: 1, h: 42, m: 2, s: 3 } });

becomes

offset 2y1d42h2m3s

Using the Expression class

The Expression class can be used to compose reusable PromQL expressions to be further used with the promql library.

new Expression({
  metric: 'test_metric',
  values: {
    arg1: 'foo',
    arg2: 'bar'
  },
  defaultOperator: MatchingOperator.regexMatch,
  defaultSelectors: [{ label: 'baz', operator: MatchingOperator.notEqual, value: '' }],
}).toString(),

becomes

test_metric{baz!="", arg1=~"foo", arg2=~"bar"}

which can then be used with a promql method

promql.max({
  by: 'baz',
  expr: new Expression({
    metric: 'test_metric',
    values: {
      arg1: 'foo',
      arg2: 'bar',
    },
    defaultOperator: MatchingOperator.regexMatch,
    defaultSelectors: [{ label: 'baz', operator: MatchingOperator.notEqual, value: '' }],
  }).toString(),
});

becomes

max by (baz) (test_metric{baz!="", arg1=~"foo", arg2=~"bar"})

see promql.ts for all available methods.