Research Spikes¶
Time-boxed research spikes that gate milestone exits. Each report ends with a go/no-go recommendation and the impact on the platform plan if no-go.
M1 exit spikes (plan §12.2 items 6–8)¶
| Spike | Report | Track | Recommendation |
|---|---|---|---|
| OpenFGA performance at target scale | m1-openfga-performance | A — real execution (local OpenFGA v1.18.3 + synthetic tuple load) | GO-WITH-CONDITIONS |
KRO v1alpha1 upgrade drill |
m1-kro-upgrade-drill | B — reproducible harness, preliminary findings | GO-WITH-CONDITIONS |
| Bundle-managed ArgoCD lifecycle | m1-argocd-bundle-lifecycle | B — reproducible harness, preliminary findings | GO-WITH-CONDITIONS |
Re-running the experiments¶
All harnesses live under harness/:
harness/openfga/— Go harness: authorization model (model.json), synthetic tuple generator at the v1 scale envelope and 10×, Check/ListObjects benchmark. Requires only a localopenfgabinary.harness/kro/run.sh— kind + kro upgrade drill (requires docker, kind, kubectl, helm).harness/argocd/run-upgrade-drill.shanddetect-byo.sh— bundle upgrade/rollback drill and BYO detection probe (requires docker, kind, kubectl).
Track B reports contain figures marked PRELIMINARY; they become final when the harness is re-run on a docker-capable machine and the artifacts are attached to the report.