5.2 KiB

Raw Permalink Blame History

Thanos Proposal

Name of project: Thanos

Description:

Thanos is a metric system that aims for being simple and cheap option to centralize and scale out the Prometheus based systems. It achieves that by:

Gradual and incremental installation model. Users can seamlessly transform vanilla Prometheus system into partial Thanos, optionally upgrade into fully fledged Thanos deployment with near unlimited retention capabilities in separate isolated steps.
Single, widely available, cheap and optional dependency: object storage
Built-in Prometheus HA support and downsampling capabilities
Flexible deployment model due to a cloud-native, microservice approach using gRPC.

Concretely the aims of the project are:

Global query view of metrics.
Unlimited retention of metrics, including downsampling.
High availability of components, including support for Prometheus HA.

Statement on alignment with CNCF mission:

The Thanos project aligns well with the mission of the CNCF by empowering organizations to build, run and monitor applications at scale.

It can be used to monitor and alert on microservices and infrastructure that spans over not only multiple clusters but over multiple clouds with a simple approach to adoption and scaling.

Thanos aims to enable people to take their current Prometheus ecosystem and scale it out beyond their current capabilities, gradually and without downtime.

Building on top of the work that is being done by the Prometheus community we can scale out to give Thanos users a global view of all their metrics without the need of aggregations.

Thanos has similar goals to Cortex project that is already a CNCF Sandbox project. However, Thanos approached the problem with the focus on simplicity and gradual installation. It takes advantage of having only single dependency (object storage) and is integrating with Prometheus in slightly different way. On the other hand Cortex is more mature and have more performance optimizations. As per No Kingmakers & One Size Does Not Fit All CNCF principle we think it makes sense to have both projects part of CNCF. Worth to note that we are already collaborating with Cortex team on reusable improvements that would benefit both projects. In fact, we believe that being together in CNCF will improve the collaboration even further.

Sponsors / Advisors from TOC: Xiang Li

Unique identifier: Thanos

Preferred maturity level: Sandbox

License: Apache License 2.0

Source control repositories: https://github.com/improbable-eng/thanos

We also own neutral github organization which we plan to use: https://github.com/thanos-io

External Dependencies: