Skip to content

Roadmap

dokimi is built in milestones. Each one delivers something usable, and nothing is built before the design it depends on is settled. Milestone-01 becomes the first release, 0.1.

Milestone-01: the first end-to-end run

in progress

A deliberately narrow first slice. One cluster, one iperf2 UDP test between two nodes you name, over the pod network, measured through a managed Prometheus, evaluated against throughput and packet-loss criteria, results written as CSV and JSONL, and everything the run created cleaned up afterwards.

Work item State
Repository, toolchain and continuous integration done
Orchestrator container image and release wiring done
Test definitions in YAML and Robot done
CLI: validate, dry-run, submit, stream, reattach done
Cluster discovery, node resolution and planning done
A multi-node test environment on one machine done
Rename to dokimi done
Run identity, ownership labels, one run at a time per cluster done
Preflight: check the cluster before running done
Managed Prometheus, with a node-exporter when the cluster has none done
The iperf2 test type: its plugin, its image and the per-test readiness gate done
Choosing every image a run uses when you submit done
The iperf2 exporter done
Running the test through Robot Framework, with Robot's report done
Bringing Robot's report back to your machine done
Measurement done
Evaluating the pass criteria done
Result files: CSV and JSONL planned
Post-run diagnostics done
Outcomes, cleanup and the final exit code done
An end-to-end acceptance run planned

After the first release

planned

These milestones are planned, not scheduled. Their order can change as the first release teaches us more.

Milestone-02: whole-cluster test matrices. From one pair of nodes to tests across a cluster: all-to-all and one-to-many matrices, selecting nodes by label or topology, running pairs in parallel where it is safe, and a versioned YAML schema.

Milestone-03: more test types. iperf3 and IRTT next to iperf2, including latency, without test-type-specific code in the core.

Milestone-04: results that last. An optional external result database, repeated runs and aggregation, historical baselines, and regression detection.

Milestone-05: stable interfaces. The interfaces move from v1alpha1 to v1, with explicit compatibility guarantees and migration rules, ahead of a Kubernetes Operator.

Milestone-06: multi-cluster. Tests between clusters: discovery, connectivity, security and correlating results from more than one cluster.