levitate-streaming-aggregation-automation

This is an example repository which stores the streaming-aggregation config, as well as aggregation pipelines to be executed on the metrics. Feel free to fork.

Read more about streaming aggregates in Levitate here.

Config file

We need to have a config file for each cluster on which we want to run out aggregations on. This is present in config.json

{
  "clusters": {
    "cluster-name": {
      "region": "cluster-region",
      "tenant": "tenant-name",
      "cluster_id": "70fb636d-11e8-7b6a-bd3c-789b991ad626"
    }
  }
}

Aggregations file

The aggregations are stored in a file called cluser-name.yaml in the root of the repository. There is one file per Levitate cluster.

How to define the streaming aggregated metric?

Update the cluser-name.yaml file with as follows:

- promql: 'sum2 by (stack, le) (http_requests_duration_seconds_bucket{service="pushnotifs"}[1m])'
  as: aggregated_http_requests_duration_seconds_bucket

promql is the definition of the aggregated metric. as is the name of the new aggregated metric available for further querying.

Supported functions

Function Name	Description
sum	Total to be used for other metric types
count	A count of the number of samples.
max	The Maximum value of the samples
sum2	Sum, but for counters and reset awareness.
increase	The increase in counter value.
min	The minimum value of the samples
avg	The average value of the samples.