When querying grouped series with non unitary sampling (eg: 2 minutes), some gapped groups returns some points that aren't aligned with ungapped groups timestamps.
This way, I'll get common HTTP codes like 200 that will be sampled every 2 minutes (generally on even ones), but some rare codes that aren't present on every minute will be sampled on the exact minute they appear, which is sometimes an odd minute, creating a timestamp that doesn't exist on full groups.
I thought using align_sampling should align different groups on the same timestamps but it seems not.
Hi,
When querying grouped series with non unitary sampling (eg: 2 minutes), some gapped groups returns some points that aren't aligned with ungapped groups timestamps.
Here's a query example:
This way, I'll get common HTTP codes like 200 that will be sampled every 2 minutes (generally on even ones), but some rare codes that aren't present on every minute will be sampled on the exact minute they appear, which is sometimes an odd minute, creating a timestamp that doesn't exist on full groups.
I thought using
align_sampling
should align different groups on the same timestamps but it seems not.Using gaps aggregator doesn't help.