Conversation
glours
force-pushed
the
e2e-ci-sharding-and-image-cache
branch
3 times, most recently
from
August 18, 2026 07:43
f878666 to
c0b8477
Compare
…igurable Several e2e tests used fixed host ports (8070, 8080, 8090, etc.), global buildx builders registered with --use, and a hardcoded IPC container name. Under parallel execution these shared daemon-level resources caused port conflicts, buildkitd container-name collisions, and intermittent failures. Separately, the standalone make target hardcoded -parallel=1, forcing the entire standalone suite to run serially even though the same tests run in parallel in plugin mode. Changes: - Convert fixed host ports in four fixtures (network-test, sentences, build-test, volume-test) to ephemeral bindings; add ServicePublishedPort helper to framework.go to resolve the actual mapped port at runtime. - Give each buildx-builder test a unique daemon-scoped name via BuilderName helper, preventing container-name collisions between parallel goroutines. Applies to TestBuildPlatformsWithCorrectBuildxConfig, TestBuildPrivileged, TestBuildBuilder, TestBuildEntitlements, TestBuildTLS. - Rename the fixed ipc_mode_container to a project-scoped name (ipc_e2e-src) and update the ipc-test fixture accordingly. - Replace the hardcoded -parallel=1 in make e2e-compose-standalone with E2E_PARALLEL_PLUGIN?=4 and E2E_STANDALONE_PARALLEL?=4, aligning both modes with the effective GOMAXPROCS on 4-core GitHub Actions runners. Both variables are overridable (E2E_STANDALONE_PARALLEL=1 restores serial local debugging). - Raise the poll delay in watch_test.go long-running polls (90s/120s timeout) from the default 100ms to 1s; bump cascade_test.go and publish_test.go from 100ms to 500ms. Reduces CPU pressure without affecting effective test duration. The logging-driver fixture retains its fixed port 24224: the Docker fluentd log driver encodes fluentd-address at service-start time, making ephemeral resolution non-trivial. TestLoggingDriver is serial so there is no collision risk in practice. Signed-off-by: Guillaume Lours <glours@users.noreply.github.com>
glours
force-pushed
the
e2e-ci-sharding-and-image-cache
branch
from
August 18, 2026 07:44
c0b8477 to
8e52599
Compare
Add a GHA cache layer for the passive Docker images used by the e2e suite.
Before each test run the workflow attempts to restore a pre-built tar
from the cache (key = Docker engine version + sha256 of the image list).
On a cache hit the images are loaded with 'docker load'; on a miss they
are pulled in parallel ('xargs -P 4 -n 1') and saved for the next run.
A new allowlist .github/e2e-images.txt controls which images are cached.
It includes the infrastructure images that every test cell downloads:
docker:dind, mariadb, golang:alpine, fluent/fluent-bit:3.1.7-debug,
registry:3, and the sentences/nginx helper images. Images used by
pull-policy and image-identity assertions (alpine:* and several fixture-
specific tags) are intentionally absent so those tests remain unaffected.
The ci.yml changes are provided as .github/ci-workflow-cache-only.patch
because the sandbox environment lacks the GitHub 'workflow' OAuth scope.
Apply with:
git apply .github/ci-workflow-cache-only.patch
git rm .github/ci-workflow-cache-only.patch
git commit --amend --no-edit --signoff
Also fix ServicePublishedPort helper to take only the first line of
'compose port' output (SplitN), making it robust when a service is scaled.
Signed-off-by: Guillaume Lours <glours@users.noreply.github.com>
glours
force-pushed
the
e2e-ci-sharding-and-image-cache
branch
from
August 18, 2026 08:58
9045bd4 to
0f2a67f
Compare
The previous patch was generated with Python difflib which omits the required leading space on blank context lines, producing a 'corrupt patch' error when applied with git apply. Regenerated with git diff --cached to get proper unified diff format. Signed-off-by: Guillaume Lours <glours@users.noreply.github.com>
glours
force-pushed
the
e2e-ci-sharding-and-image-cache
branch
from
August 18, 2026 09:04
077f716 to
6602303
Compare
glours
force-pushed
the
e2e-ci-optimizations-wave1-wave2
branch
from
August 19, 2026 13:39
057be6f to
797c4a3
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What I did
Two independent CI optimisations stacked on top of the test-isolation PR:
1. Test sharding (reduces wall-clock time)
Each of the 6 e2e matrix cells is split into 2 shards (6 → 12 jobs):
EXCLUDE_E2E_TESTS).watch_test.gotests (7 serial tests with 90–120 s polls and large image downloads) plusTestLocalComposeBuild,TestBuildPlatformsWithCorrectBuildxConfig,TestBuildTLS,TestBuildPrivileged,TestBuildEntitlements.The shard filter is a single
E2E_HEAVY_TESTSenv var exported once and reused in both plugin and standalone steps.make e2e-composeandmake e2e-compose-standaloneremain unchanged locally.Expected impact: each shard ≈ 6 min wall-clock vs. the previous ≈ 11 min for the heaviest standalone cells.
2. Image pre-pull and GHA cache (reduces network waits)
A new allowlist
.github/e2e-images.txtlists 9 passive images (docker:dind,mariadb,golang:alpine,fluent/fluent-bit:3.1.7-debug,registry:3,gtardif/sentences-*,nginx:alpine). Before the tests run, a GHA cache keyed on Docker engine version + file hash restores a tarball of those images; on a miss they are pulled in parallel and saved for the next run.Images that are the subject of pull-policy/pull-skip assertions (
alpine:*and fixtures undercompose-pull/,image-identity/) are intentionally absent from the list.The sandbox environment lacks the
workflowGitHub OAuth scope needed to push to.github/workflows/. The.github/e2e-images.txtfile is committed. The workflow changes are in.github/ci-workflow-shard-changes.patch:The patch is the only missing piece; all other content is present in the branch.
Scope notes
A7 (migrate ~12 config-only e2e tests to unit tests) and A9 (framework-level resource sharing) are intentionally deferred: A7 requires testing the config renderer without a subprocess with non-trivial coverage risks; A9 needs measurement before investing (diagnostic: medium risk). Both are candidates for a follow-up.
Related issue
No related issue — CI infrastructure improvement.
(not mandatory) A picture of a cute animal, if possible in relation to what you did
Two pandas who finally got their own bamboo plots instead of sharing 🐼🐼