You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As explained on Slack, when we compute the isGoodSlice field on the rollup document, we set it to 0 whenever there is no events during that bucket. This makes the SLO tank anytime we don't have data for some period of time, which is not coherent with how the occurrences budgeting method interprets no data, i.e. sum(events) + 0 = sum(events).
Ideally, the outcome of a slice should not be a boolean (isGoodSlice) but one of: good, bad, no_data. Doing this would require some migration and breaking change that we can't deal with at the moment.
By changing isGoodSlice to 1 whenever there is no data, we would compute the SLI as: 1 - (total - good) / total slices in time window = (1 - (value_count_agg(isGoodSlice) - sum_agg(isGoodSlice)) / total slices in time window)`
Ideally, we would compute the sli as: sum_agg(slice = good) / total slices in time window
The text was updated successfully, but these errors were encountered:
🍒 Summary
As explained on Slack, when we compute the isGoodSlice field on the rollup document, we set it to 0 whenever there is no events during that bucket. This makes the SLO tank anytime we don't have data for some period of time, which is not coherent with how the occurrences budgeting method interprets no data, i.e.
sum(events) + 0 = sum(events)
.Ideally, the outcome of a slice should not be a boolean (isGoodSlice) but one of: good, bad, no_data. Doing this would require some migration and breaking change that we can't deal with at the moment.
By changing isGoodSlice to 1 whenever there is no data, we would compute the SLI as:
1 - (total - good) / total slices in time window
= (1 - (value_count_agg(isGoodSlice) - sum_agg(isGoodSlice)) / total slices in time window)`Ideally, we would compute the sli as:
sum_agg(slice = good) / total slices in time window
The text was updated successfully, but these errors were encountered: