Migration notes for SignalFx (Splunk) to PromQL
There are several key differences between SignalFx (Splunk) and PromQL of which to be aware when migrating queries to Chronosphere:
- Rollups in SignalFx are used to aggregate data points within a single time series. With Chronosphere, rollups are defined only on ingest, and not at query time. However, PromQL offers a set of functions (opens in a new tab) for aggregating individual time series over time.
- Analytics in SignalFx are the equivalent to functions in PromQL.
Equivalent function list
| SignalFx | PromQL | Notes |
|---|---|---|
| Absolute value | abs() | none |
| Bottom | bottomk() | none |
| Ceiling | ceil() | none |
| Count | count(), count_over_time() | none |
| Delta | delta() | none |
| EWMA and Double EWMA | holt_winters() | holt_winters() is similar to, but not exactly like, Exponentially Weighted Moving Average |
| Exclude | Use comparators | none |
| Floor | floor() | none |
| Integrate | rate() | none |
| LN - Log (natural) | ln() | none |
| Log10 | log10() | none |
| Maximum | max(), max_over_time() | none |
| Mean | avg(), avg_over_time() | none |
| Minimum | min(), min_over_time() | none |
| Percentile | quantile(), quantile_over_time() | Use histogram_quantile() with Histograms |
| Power | Use operators | none |
| Rate of change | rate() | none |
| Scale | Use operators | none |
| Square root | sqrt() | none |
| Standard deviation | stddev(), stddev_over_time() | none |
| Sum | sum(), sum_over_time() | none |
| Timeshift | offset | none |
| Top | topk() | none |
| Variance | stdvar(), stdvar_over_time() | none |