Adoption & Ecosystem
Who uses it
The organizations below are named with citable sources. The first five were cited by the CNCF at the time of Thanos's incubation; Wikimedia publishes its own operations documentation.
| Organisation | Use case | Source |
|---|---|---|
| Alibaba Cloud | Production metrics | CNCF blog, 2020-08-19 |
| Banzai Cloud | Production metrics | CNCF blog, 2020-08-19 |
| HelloFresh | Production metrics | CNCF blog, 2020-08-19 |
| Monzo | Production metrics | CNCF blog, 2020-08-19 |
| Red Hat | Production metrics | CNCF blog, 2020-08-19 |
| Wikimedia | Long-term Prometheus storage | Wikitech: Thanos |
Adoption signals
Measured on 2026-06-25 via the GitHub API on thanos-io/thanos:
- Stars: 14,124
- Forks: 2,318
- Open issues: 869
- Contributors: roughly 408 (last page of the paginated contributors list)
Releases land roughly every six weeks as a single binary on GitHub Releases plus the quay.io/thanos/thanos container image.
Ecosystem
- Prometheus: the data source, via the sidecar or remote-write into Receive.
- Grafana: queries Thanos through the Prometheus-compatible query API.
- Deployment: Prometheus Operator and Helm charts.
- Object storage: S3, GCS, Azure, Swift, and Tencent COS through the
thanos-io/objstorelibrary. - Learning: interactive Killercoda tutorials.
Alternatives
Cortex entered CNCF incubation on the same day as Thanos (see the CNCF blog, 2020-08-19). VictoriaMetrics and M3 are the other common comparisons.
| Alternative | Differs by |
|---|---|
| Cortex | Push and multi-tenant aggregation from the start, rather than sidecar-first retrofit onto existing Prometheus. |
| VictoriaMetrics | Its own storage engine sold on single-system performance, instead of reusing the TSDB block format on object storage. |
| M3 | Distributed time-series store with its own clustering model. |
Thanos's distinguishing point is that it bolts onto an existing Prometheus with a sidecar, places the TSDB block format directly in object storage, and uses the recursive StoreAPI abstraction plus downsampling to make global and hierarchical queries work. Push-based aggregation via Receive came later as a second option.