e2e/

qcloud e2e tests

End-to-end tests that run the real qcloud binary against a real Qdrant Cloud backend. These tests create paid resources on whatever account you point them at — never run them against a production account without first double-checking the credentials.

Running

make e2e

Or directly:

QCLOUD_E2E=1 \
QDRANT_CLOUD_API_KEY=... \
QDRANT_CLOUD_ACCOUNT_ID=... \
go test -timeout 20m -v -count=1 ./test/e2e/...

Without QCLOUD_E2E=1 every test in this tree is skipped, so go test ./... from the repo root stays network-free.

Environment variables

Binary acquisition

The binary is resolved exactly once per go test invocation in TestMain:

1. If QCLOUD_E2E_BINARY is set, it is used as-is.
2. Otherwise the qcloud-<os>-<arch>.tar.gz archive from the release given by QCLOUD_E2E_RELEASE (default latest) is downloaded, its sha256 is checked against the release's checksums.txt, and the binary is extracted into $XDG_CACHE_HOME/qcloud-e2e/<sha>/qcloud (or the platform equivalent).

Cache entries are keyed by the archive's sha256, so repeated runs on the same host reuse the same extracted binary — and automatically pick up a fresh copy when the upstream release changes.

To test a locally-built binary:

make build
QCLOUD_E2E_BINARY=$PWD/build/qcloud make e2e

Writing new tests

Every test starts with framework.NewEnv(t); it returns an *Env that wraps the binary path, an isolated empty config file, and generic CLI wrappers:

• env.Run(t, args...) — runs qcloud, fails the test on non-zero exit, streams stdout/stderr to t.Log.
• env.RunAllowFail(t, args...) — same, but returns the result instead of failing. Use it for negative tests.
• env.RunJSON(t, &v, args...) — appends --json and decodes stdout into v.

Resource-specific helpers live next to the tests that use them — they're ordinary _test.go files in package e2e_test, free to grow without bloating the framework. For clusters, see cluster_helpers_test.go:

• createCluster(t, env, opts) — creates a cluster and registers cleanup. The cluster name defaults to e2e-<random>, which makes leak sweeps safe.
• waitCluster(t, env, id, timeout) — blocks until the cluster is healthy.
• sweepLeakedClusters(t, env, maxAge) — best-effort cleanup of stale e2e-* clusters; call it from a dedicated test if you want automatic housekeeping in CI.

Prefer RunJSON over scraping human output — the JSON shape is stable, human output is not.

Safety notes

• Every cluster created through createCluster is scheduled for deletion via t.Cleanup. If a runner is killed mid-test, orphans remain; use sweepLeakedClusters to catch them.
• Tests are not safe to run with t.Parallel() today — they share a single account and a small quota. Don't add t.Parallel() without also isolating each test's account or region.