Allow users to create native histograms#1104
Conversation
csmarchbanks
force-pushed
the
build-native-histograms
branch
from
d9708f3 to
7f01fd3
Compare
csmarchbanks
force-pushed
the
build-native-histograms
branch
from
7f01fd3 to
8f5698b
Compare
Initial support for allowing users to create native histograms. The initial implementation directly exposes the initial schema version rather than a factor, and we can provide guidance for choosing a good initial schema. If too many buckets will be created then resolution can be reduced, but in the current implementation the other reduction techniques (increasing zero bucket width, resets, etc...) are not yet implemented. Signed-off-by: Chris Marchbanks <csmarchbanks@gmail.com>
csmarchbanks
force-pushed
the
build-native-histograms
branch
from
8f5698b to
64414de
Compare
|
@krajorama Would you be willing to take a look at this sometime? I tried to describe the differences I made compared to say client_golang, but let me know if you have questions! |
|
Are there plans to merge this fairly soon? I'm very new to Prometheus but identified this as a pretty high-priority feature for some of our Python services. |
|
Hello! Since this PR was originally opened there have been some additional changes to the text based format. Just recently the draft of OpenMetrics 2.0 was published (https://prometheus.io/docs/specs/om/open_metrics_spec_2_0/). I need to update this client to support the changes that were made in 2.0 and then can come back to this. In addition, for it to actually be useful Prometheus/scraping libraries will need to support OpenMetrics 2.0, I am not 100% sure on the timeline for that. TL;DR This work was recently unblocked and I will come back to it soon. |
2 participants
Initial support for allowing users to create native histograms. The initial implementation directly exposes the initial schema version rather than a factor, and we can provide guidance for choosing a good initial schema. If too many buckets will be created then resolution can be reduced, but in the current implementation the other reduction techniques (increasing zero bucket width, resets, etc...) are not yet implemented.
Depends on #1087 for the actual exposition
Part of #918