Repository navigation
Conversation
Signed-off-by: Bartek Plotka <bwplotka@gmail.com>
106b612 to
622ec29
Compare
Co-authored-by: Bartlomiej Plotka <bwplotka@gmail.com> Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com>
Signed-off-by: Bartek Plotka <bwplotka@gmail.com>
|
|
||
| ## Alternatives | ||
|
|
||
| ### 1. Separate repository and binary under Prometheus governance |
There was a problem hiding this comment.
Thinking through this more I'd prefer a particular combined version of alternatives 1 and 2.
Whereby,
- Prometheus does govern scalable LTS scope, but through shared library crates (promql-rs, scrape.rs, sth sth storage.rs/tsdb.rs and friends...think anything that can be treated as reusable library, that is maybe currently hidden away as a go package in prometheus or thanos)
- prometheus/prometheus.rs composes those shared crates into a single node monolith binary
- thanos-io/thanos.rs (or corthamir.rs) composes the same as distributed infra binaries. I'd propose shared governance across orgs here even, so all parties can collaborate
This has quite a few benefits,
- From Prometheus end user pov, they still get the single binary they know and love, maybe even with prometheus-operator support, just way more efficient.
- From Thanos/Cortex/Mimir end user pov, they can gradually swap out pieces of their infra with new columnar one, maybe querier first, then ingester, then stores, as we build and create the crates for those pieces and integrate into corthamir binary.
- Since everything is maintained as shared libraries, operators who had already created their own implementations internally, can simply refactor and adopt libraries, and contribute, without having to commit to prometheus/corthamir binary. Particularly relevant for AI era. Essentially prometheus org expands governance scope to maintain building blocks for scalable LTS.
This way we also don't have to part with Prometheus/Thanos branding/seo. Wdyt?
There was a problem hiding this comment.
Nice idea! I will add as another alternative, let's discuss!
I think we should aim higher and simplify life of everyone in CNCF. We should "move" Cortex and Thanos projects into Prometheus governance effectively. I am pretty sure this is what everyone wanted from day 1 (WDYT @caniszczyk , any advice?)
Moving under a single governance has multiple benefits:
- It simplifies the already complex CNCF landscape. I think it's important.
- It gives a middle ground to soften any previous misalignments (if there were any) between Cortex and Thanos.
- It increases chances Grafana will join one day (we want that IMO)
- We don't need to do (it's difficult!) work to setup/change governance in Thanos. I personally won't be able to this work if it's not part of Prometheus.
So I think it's important it's in Prometheus, if Prometheus SC and community allows it.
There was a problem hiding this comment.
As per repo, I don't think it matters much, let's just choose one thing:
A) prometheus/prometheus.rs with multiple crates (and monolith binary)
B) prometheus/prometheus.rs with monolith binary, creates in separate small repositories
C) prometheus/prometheus.rs with monolith binary, prometheus/distro_prometheus.rs for other distributed creates
🤷🏽
There was a problem hiding this comment.
addressed in 174079e: added the shared-crates plus separate-org distributed binaries option as Alternative 2.
| ## Bringing scalable LTS scope to Prometheus governance | ||
|
|
||
| * **Owners:** | ||
| * Bartek Plotka [@bwplotka](https://github.com/bwplotka) |
There was a problem hiding this comment.
This list is based on the most active people on the SIG meeting yesterday. If you want to co-own, let me know!
Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com>
Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com>
metalmatze
left a comment
There was a problem hiding this comment.
Thanks for writing this up so quickly after yesterday's SIG meeting, @bwplotka!
One thing I tripped over is that "LTS" is ambiguous. When I first read the title, I thought of Prometheus LTS as in Long Term Support, since we already ship LTS releases. I'm sure others will read it the same way.
Long-term storage also isn't what's new here. You can already run a single Prometheus with 10 years of retention today. What actually expands the scope is horizontal scalability, so I'd suggest we drop "LTS" and frame the proposal around horizontally scalable Prometheus instead.
I also don't see why the title is about governance. From my point of view, this proposal is about bringing a horizontally scalable Prometheus written in Rust to the Prometheus org. Of course the org is under Prometheus governance, but that's an implementation detail. Something like "proposal: PROM-116: Horizontally scalable Prometheus in Rust" would say that more directly.
friedrichg
left a comment
There was a problem hiding this comment.
Fully in support. Only one minor nit
Signed-off-by: Bartek Plotka <bwplotka@gmail.com>
… scalable Prometheus in Rust Signed-off-by: Bartek Plotka <bwplotka@gmail.com>
…onflict Signed-off-by: Bartek Plotka <bwplotka@gmail.com>
…ernative Signed-off-by: Bartek Plotka <bwplotka@gmail.com>
|
Great comments, changed "LTS" to a new title, thanks! |
Potentially controversial one, but let's see. We always wanted to do this, but now timing might be great 💪🏽
Following up on the November 19, 2026 Prometheus.rs SIG discussion, this proposal brings horizontally scalable, distributed long-term storage requirements into official Prometheus governance for the next-generation Rust and columnar metrics effort.
Key areas for feedback from maintainers and the Steering Committee: