Increase indexheader_lazy_load_duration_seconds buckets

[pracucci]

• I added CHANGELOG entry for this change.
• Change is not relevant to the end user.

Changes

In Cortex we've noticed that the 99th percentile computed onindexheader_lazy_load_duration_seconds is always capped to 5s even if the average can be way higher. Reason is that the largest histogram bucket is 5s but lazy loading can take way longer than that (mostly depending on index size and the disk speed).

In this PR I proposes to add more buckets to the metric, to have a better estimation of 99th percentile on high durations.

Change is not relevant to the end user.

Is it? Do you track such changes in CHANGELOG?

Verification

N/A

[squat]

Thanks @pracucci

[yeya24]

Thanks @pracucci. I think 300s is little bit too large but overall LGTM.

[pracucci]

Thanks @pracucci. I think 300s is little bit too large but overall LGTM.

We've actually seen several minutes too, on a store-gateway under heavy load with indexes each of few tens GB each.

[pracucci]

If there are no concerns, can we get this merged, please?

[squat]

thanks @pracucci 🌮

[pracucci]

Thanks to you @squat ! 🌮