[Bug]: UT coverage Aliyun object-storage tests fail on endpoint timeouts

[gouhongshen]

Is there an existing issue for the same bug?

• I have checked open and closed issues for the failing test name, endpoint, package, and CI job.

Branch Name

main (PR target)

Commit ID

7ee5091

Other Environment Information

• Workflow: MatrixOne ALL CI
• Job: Matrixone UT Coverage / UT Coverage on Ubuntu/x86
• Workflow run (attempt 1): https://github.com/matrixorigin/matrixone/actions/runs/30667879680
• Failed job: https://github.com/matrixorigin/matrixone/actions/runs/30667879680/job/91279105172
• Diagnostic artifact: https://github.com/matrixorigin/matrixone/actions/runs/30667879680/artifacts/8808069956

Actual Behavior

The UT coverage job failed only in github.com/matrixorigin/matrixone/pkg/fileservice. The TestObjectStorages/aliyun/* variants timed out while connecting to the external endpoint mo-ci.oss-cn-guangzhou.aliyuncs.com (219.133.47.12:443). Representative errors from the diagnostic artifact are:

Get "https://mo-ci.oss-cn-guangzhou.aliyuncs.com/?bucketInfo": dial tcp 219.133.47.12:443: i/o timeout
Head "https://mo-ci.oss-cn-guangzhou.aliyuncs.com/...": dial tcp 219.133.47.12:443: i/o timeout
HeadBucket: request canceled, context deadline exceeded

There were no build failures. Multiple Aliyun SDK, MinIO, and AWS-wrapper subtests failed at the same external endpoint.

This occurred on PR #26540, but that PR changes only frontend data-branch output logic, its focused unit test, and its BVT SQL/result. It does not change pkg/fileservice. The immediately preceding coverage job on the identical commit succeeded: https://github.com/matrixorigin/matrixone/actions/runs/30665245694/job/91270680472

Expected Behavior

The Aliyun object-storage tests should complete reliably without intermittent CI failures caused by inability to reach their external test endpoint.

Steps to Reproduce

1. Run the MatrixOne ALL CI UT coverage job on commit 7ee509137922722402c7738f7c4bb1c0d6753fd3.
2. Allow pkg/fileservice integration tests to access the configured Aliyun test endpoint.
3. Observe that the failure is nondeterministic: the prior run on the same commit passed, while run 30667879680 timed out connecting to the endpoint.

Additional information

The failure is external-network/environment related rather than caused by the PR diff. This issue tracks the flaky endpoint dependency; the unrelated failed workflow will be rerun once according to the PR CI handling policy.

[gouhongshen]

Root cause confirmed and fix submitted.

The failed coverage run never reached HTTP: the Tencent Guangzhou TKE runner timed out establishing TCP to Aliyun OSS. The same commit passed on rerun from another ephemeral runner, so credentials and SDK request behavior are not the direct cause. The structural defect is that go test -short discovered ambient TEST_S3FS_* credentials and ran real cloud integration tests inside required PR coverage.

Fix:

• ci: separate object storage integration from coverage UT CI#416 removes cloud credentials from coverage UT and provides provider-isolated integration jobs that can test an exact PR head.
• test(fileservice): isolate external object storage integration #26612 separates local and external fileservice tests. PRs changing pkg/fileservice, go.mod, or go.sum still run real Aliyun/QCloud integration before merge; unrelated PRs do not. Daily/manual runs continue detecting provider and network drift.

Merge order: matrixorigin/CI#416 first, then #26612. Both PRs remain Draft while CI runs.

[gouhongshen]

Automated AI triage

This issue requires human involvement and cannot currently be completed independently by the
automated fix loop.

Reason

Requires object-storage/CI infrastructure ownership; next action: restore reliable access to the Aliyun endpoint or provide a maintained isolated test service.

Evidence

Multiple storage-client variants timed out against the same external OSS endpoint; an immediately preceding run on the same commit succeeded.

Please use this analysis to provide the missing decision, environment, evidence, or ownership
needed before the issue is returned to the AI candidate pool.