End-to-end CI scripts

Orchestrator scripts for the Semaphore e2e jobs. Two top-level entry points drive two different job shapes:

Entry point	Job shape
`body_standard.sh`	Standard e2e: provision a cluster, install Calico, optionally migrate/upgrade, run tests
`body_flannel-migration.sh`	Flannel-to-Calico migration test with a pre- and post-migration test run

Both dispatch to single-purpose phase scripts under phases/. Each phase is self-contained, documents its required env vars at the top, and can be sourced individually when reproducing part of a CI run locally.

Phases

Phase	Purpose
`phases/provision.sh`	`bz provision` + Semaphore cache store
`phases/install.sh`	`bz install` (install Calico on the provisioned cluster)
`phases/configure.sh`	Post-install env setup: PATH, external-node creds, IPAM pool, failsafe ports
`phases/migrate.sh`	Optional operator migration, AKS migration, `bz upgrade`
`phases/load_images.sh`	Side-load PR-built helper images (rapidclient) onto nodes; no-op except on the local-binary gcp-kubeadm PR path
`phases/run_tests.sh`	Acquire and run the e2e binary (local build, hashrelease download, or `bz tests` fallback)
`phases/hcp.sh`	Hosted control plane flow (separate provision + test tooling)

Reproducing a CI run locally

Each phase script lists its required env vars in its header comment. In the common case, reproducing a CI job looks like:

bash

cd "${BZ_HOME}"
source phases/provision.sh
source phases/install.sh
source phases/configure.sh
source phases/run_tests.sh

Phases are sourced, not executed, so env vars exported by earlier phases (e.g. PATH, EXT_IP) flow into later phases. Running a phase standalone works the same way -- source it from a shell you've set up with the required env vars.

Adding a new phase

Create phases/<name>.sh with a header comment listing required env vars.
Omit set -eo pipefail from the phase -- the orchestrator sets it once and phases inherit via sourcing.
Add the phase to the appropriate body script's dispatch logic.
Add a row to the phase table above.

The test runner

phases/run_tests.sh selects the test execution strategy automatically:

Condition	Strategy
`RUN_LOCAL_TESTS` is set	Build the e2e binary from local source (per-PR CI)
`TEST_TYPE == k8s-e2e`	Download the pre-built binary from the hashrelease (scheduled CI)
Otherwise	Fall back to `bz tests` (benchmarks, certification, etc.)

The first two paths run the binary via make e2e-run inside calico/go-build. Developers can use the same target directly:

bash

KUBECONFIG=/path/to/kubeconfig \
  E2E_TEST_CONFIG=e2e/config/gcp-bpf.yaml \
  make e2e-run

See e2e/config/*.yaml for available test-selection configs and e2e/pkg/testconfig/ for the config format.

Legacy notes

body_flannel-migration.sh still uses ./bz.sh tests:run for its pre- and post-migration test runs -- that's a different legacy runner than bz tests and has tests the in-repo binary doesn't yet cover. Migrate to make e2e-run when parity lands.