AlinsRan opened a new pull request, #2788:
URL: https://github.com/apache/apisix-ingress-controller/pull/2788
## Description
This is the first ("P0") step of an e2e test runtime/stability optimization
effort. It contains only low-risk, mechanical changes that live entirely in the
`test/e2e` scaffold/framework layer — **no test spec bodies are modified**, so
the changes are provider-agnostic and equally apply to downstream provider
implementations of the `scaffold.Deployer` interface.
## Changes
- **Drop the hardcoded `time.Sleep(3s)` in `APISIXDeployer.AfterEach`.**
Every spec uses a unique namespace, so the inter-spec pause only padded
wall-clock time. Removing it saves ~3s per spec.
- **Delete the test namespace with `--wait=false`** so teardown no longer
blocks on slow namespace finalization.
- **Set `imagePullPolicy: IfNotPresent`** on the dev `apisix` /
`apisix-ingress-controller` / `adc` images to avoid sporadic registry pulls
inside kind (a flakiness + latency source).
- **Make `DefaultTimeout` / `DefaultInterval` overridable** via
`E2E_DEFAULT_TIMEOUT` / `E2E_DEFAULT_INTERVAL` env vars (Go duration syntax),
so CI can tune retry budgets under high concurrency without code changes.
## Notes / things to watch in CI
Removing the 3s inter-spec pause means parallel ginkgo nodes start deploying
pods more densely; the non-blocking namespace delete partially offsets the
cleanup contention. This PR is intentionally small so the real CI run can
validate peak-load behavior.
## Follow-ups (not in this PR)
- P1: convert the ~190 `time.Sleep` calls in spec bodies to
`Eventually`/`RetryAssertion` polling (the main flakiness source).
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
