mirror of
https://github.com/milvus-io/milvus.git
synced 2026-10-28 17:05:53 +08:00
zhuwenxing
3e7b80b5f0
The post-chaos collection verification stage can exceed its 60-minute timeout because each concurrent worker creates a dedicated recovery continuation collection. With five concurrent workers and one single-request workload, verification includes five ordinary collections and six recovery collections. At three workers and at least 16 minutes per case, this requires at least 64 minutes before other overhead. Keep recovery collection creation in the single-request workload and remove duplicate creation from concurrent workers. The concurrent workload still resumes existing recovery collections and verifies their retained input journals. For this job configuration, post-chaos verification is reduced from 11 cases to 6; recovery continuation coverage uses one collection instead of six. Validation: - Ruff lint and formatting checks passed. - Python compilation passed. - Isolated initializer checks with mocked checker constructors passed: five concurrent workers plus one single-request worker create one recovery checker; existing recovery collections resume exclusively; ordinary collections create no recovery checker. - Jenkins/E2E was not rerun. Actual runtime improvement remains unverified. Signed-off-by: zhuwenxing <wenxing.zhu@zilliz.com>