Rate Limits

A rate limit is a strategy for limiting the usage of a shared resource across parallel components in a Wombat instance, or potentially across multiple instances. They are configured as a resource:

rate_limit_resources:
  - label: foobar
    local:
      count: 500
      interval: 1s

And most components that hit external services have a field rate_limit for specifying a rate limit resource to use, identified by the label field. For example, if we wanted to use our foobar rate limit with an http_client input it would look like this:

input:
  http_client:
    url: TODO
    verb: GET
    rate_limit: foobar

By using a rate limit in this way we can guarantee that our input will only poll our HTTP source at the rate of 500 requests per second.

Some components don’t have a rate_limit field but we might still wish to throttle them by a rate limit, in which case we can use the rate_limit processor that applies back pressure to a processing pipeline when the limit is reached. For example, if we wished to limit the consumption of lines of a csv input to a specified rate limit we can do that with the following:

input:
  csv:
    paths:
      - ./foo.csv
  processors:
    - rate_limit:
        resource: foobar

Metrics

The following metrics are exposed for a rate limit:

rate_limit_checked: A count of the number of times the rate limit has been probed.
rate_limit_triggered: A count of the number of times the rate limit has been triggered by a probe.
rate_limit_error: A count of the number of times the rate limit has errored when probed.