Describe the bug

I have hit a strange situation when tinkering with slo-exporter for the first time. I suspect processMatrixResultAsIncrease calculates prometheus metric increases incorrectly when the query interval is shorter than the interval data is updated at prometheus. I have been manually replaying queries reported by the debug-level log and I get no result from prometheus most of the time. On the other hand I run a loop that hits the foo service with several requests each second. Strangely I see no events emitted even when prometheus finally updates the values.

I suspect this assignment erases any previously collected samples when prometheus answers with no results. When the ingester finally collects something then there is nothing to compare it to. I suppose the correct way to update previousResult would be to merge it somehow, e.g.:

func (r *queryResult) update(newResult queryResult)  {
	r.timestamp = newResult.timestamp

	for k, v := range newResult.metrics {
		r.metrics[k] = v
	}
}

How to reproduce the bug

I run prometheusIngester with these settings

modules:
  prometheusIngester:
    apiUrl: "http://127.0.0.1:9090"
    queryTimeout: 30s
    queries:
      - type: counter_increase
        query: 'istio_requests_total{destination_canonical_service="foo"}'
        interval: 5s

Then I saw a neverending list of messages like this:

time="2021-10-06T18:02:42.269+02:00" level=debug msg="executed query" component=prometheusIngester duration=7.28613ms query="istio_requests_total{destination_canonical_service=\"foo\"}[5s]" timestamp="2021-10-06 18:02:42.261701125 +0200 CEST m=+180.026099376"

Expected behavior

I expect messages from sloEventProducer etc.

Additional context

None