prisma-risk / tsoracle / 26408577353

95%

push

github

test(e2e): kind-based multi-node cluster lane (rollout steps 2 + 3) (#410)

* test(e2e): scaffold the kube-e2e-driver crate

* test(e2e): add the monotonicity/error tracker

* test(e2e): implement cold-start and soak driver modes

* test(e2e): add the e2e driver container image

* test(e2e): add cold-start and soak driver Job manifests

* test(e2e): add the assertion-orchestration script

* ci(e2e): add the manual kube-e2e workflow

* docs(e2e): document the implemented kube lane and close the SIGTERM gap note

* test(e2e): reconcile StatefulSet SIGTERM comment with #406

* test(e2e): fix cluster image build (protoc dep + over-broad dockerignore)

The e2e/kube cluster image shipped in #405 as an unbuilt sketch. A real build
surfaces two failures: the Dockerfile.dockerignore excluded **/*.md and docs/,
stripping the include_str! files tsoracle-core needs at compile time; and the
build stage never installed protobuf-compiler, which tsoracle-proto's build
script requires. Both fixed; the cluster image now builds.

* test(e2e): point the soak Job at all replica endpoints

A real tsoracle client is configured with every replica (leader-hint failover
assumes it knows them all), so the soak driver should be too — it gives the
client a known-live peer to fail over to instead of the single Service VIP.

51 of 118 new or added lines in 2 files covered. (43.22%)

9320 of 9808 relevant lines covered (95.02%)

620078.37 hits per line