markhalonen / Smelt.jl

Julia client for Smelt.jl

Smelt.jl

Julia client for Smelt.jl

About Smelt

Smelt helps software development teams monitor their sensitive resources. These resources can include runtime, memory, api usage, function calls, and the like.

Smelt provides a simple API:

• name: The name of your metric
• value: The value of the metric
• (Optional) write: Defaults to false, determines if data is saved permanently to Smelt
• (Optional) date: The date of the metric (defaults to now)

Examples

Tracking runtime

Source Code

function performance_function()
# Lots of code, we want to know when performance changes. And we want history of performance
end

Test Code

function test_performance_function()
  start_time = now()
  performance_function()
  end_time = now()
  result = Smelt.send("Performance function runtime", end_time - start_time)
  if result.isabnormal
    fail("Smelt detected an abnormality! See $(result.link) for details")
  end
end

Tracking usage of an expensive api

Source Code

function call_expensive_api() end
end

function do_complex_workflow()
...
# Lots of code that calls `call_expensive_api` some number of times, which we'd like to track
end

Test Code

We're going to mock call_expensive_api and count the number of times it is called, and log to Smelt

function test_expensive_api_usage_in_complex_workflow()
  api_mock = mock(call_expensive_api)
  do_complex_workflow()
  result = Smelt.send("Expensive API Calls From Complex Workflow", api_mock.call_count)
  if result.isabnormal
    fail("Smelt detected an abnormality! See $(result.link) for details.")
  end
end

Things to consider

• Data shouldn't always get saved, for example running locally or from a PR.
• Could make it aware of the hardware it's running on.

Steps a user needs to take:

1. Register account and get API key
2. Include open-source client library to project
3. Write tests
4. Intelligently confiure when data should get written permanently to Smelt. It should only write when it's the master branch running on the build server.
5. Visit the Smelt web app and configure anomaly thresholds