Roadmap

[robert-milan]

Roadmap

Note: this is the roadmap covering the entire "Graphite Platform" at Grafana Labs. It includes metrictank, graphite, tsdb-gw, carbon-relay-ng and hosted-metrics-api.

It is indicative, incomplete and subject to change.

General

• Sharding by tags, more consistent partitioning distribution - partitioning metrics by nameWithTags (aka shard by tag) #1123
• Meta tags - meta-tags (previously known as extrinsic tags) #660
• Stats alongside query responses - Return measurements to users in HTTP response #1130
• Rollup indicator in Grafana -> return metadata with render responses - Enhancement - indication of returned data is rollup data grafana#11700
• Refactor cluster migration
• Dot munging overhaul - graphite dot munging overhaul #1008
• Formalize tag format. Shared validation library - Characters supported in tag values #1216
• Index reporting (break down of series counts by subpath)
• Opentracing sampling, on-demand tracing of requests
• Automated testing harness (much more advanced CI with performance testing)
• Kafka TLS and Auth support - support kafka TLS and auth #1277
• Slow query log
• RAM reductions (stream chunks instead of point slices, jit decoding, etc)
• Graphite render api 100% native - mission: "no longer require graphite-web" #1227
• Fix Stats issue - GC resulting in missed stats #1207
• Graphite response codes fix - bad requests result in http500 graphite-project/graphite-web#2461
• non-kubernetes deployments
• clustering redesign
• management UI

Carbon-relay-ng

• Agent v2 (with insights into config, support to push config updates, feedback to customers about perf issues, HA without out-of-order issues) - ideas for crng v2 carbon-relay-ng#347
• Carbon-relay-ng counter aggregator - better support for aggregation of counters carbon-relay-ng#344
• Carbon-relay-ng fix aggregator unreliable under cpu load - once a certain amount of cpu is reached, [aggregator] output becomes unreliable carbon-relay-ng#241
• Carbon-relay-ng test all packages / packaging bugs - test all of our packages carbon-relay-ng#348

tsdb-gw

• Rate limits on inbound rate - https://github.com/raintank/tsdb-gw/issues/134
• Rate limits on active series and query workload

Documentation

• Platform functionality, api's (ingestion, querying, metrics management) (cloud)
• Administrator Guide (on-prem)

[wlargou]

For new deployment, do you recommend to go with MetricTank as backend of Graphite ? And what's the roadmap for the two platforms, is there plans for merging them sometime in the near future ?

[Dieterbe]

There are no plans to merge graphite and metrictank. they are separate projects. if you want run a large graphite installation, then metrictank is a good/scaleable way to do so. but for small installations it's overkill. you can also look at Grafana Cloud and then you don't need to run anything yourself (and there's a free tier)

[wlargou]

Thanks for your feedback, the customer opted for Prometheus/Grafana, I guess that will be enough for his mid-sized environment.