flake: TestSupportBundle (multiple variants): health report should not be empty

[mafredri]

Reopening due to recurrence on 2025-11-17 in Nightly Gauntlet.

Recent failure evidence

• CI run: https://github.com/coder/coder/actions/runs/19418075437
• Job: test-go-pg (macos-latest)
• Package: github.com/coder/coder/v2/cli
• Failing subtest: TestSupportBundle/NoAgent
• Commit: 897286f33520f45e92f37c452c987124231319f2 (author: Ethan)

Key log excerpts

=== FAIL: cli TestSupportBundle/NoAgent (63.30s)
... GET /api/v2/debug/health ... took=1m0.00276575s  status_code=503
response_body={"message":"Healthcheck is in progress and did not complete in time. Try again in a few seconds."}
... support_test.go:243 ... support_test.go:153
Error: Should NOT be empty, but was {...}
Test: TestSupportBundle/NoAgent
Messages: health report should not be empty

Root-cause classification

• Flaky test (health check endpoint didn’t return within test timeout; bundle health.json ends up empty).

Precise assignment analysis

• Primary: Test function blame (approx.): recent meaningful changes to this test family:
  • chore(cli): increase healthcheck timeout in TestSupportbundle — a8fbe71a22fdc9dfd607d3ebc93c0d469cede735 (Cian Johnston)
  • test(cli): fix context init in TestSupportBundle — 860d17ad09d40a5a837013faf64e4c790c4a13ee (Mathias Fredriksson)
  • Numerous support bundle test additions/maintenance by Cian across 2024–2025.
• Given the changes above, assigning to the test owner/primary maintainer of support bundle tests.

Related issues

• This issue originally tracked NoWorkspace variant; the new occurrence is NoAgent variant. Keeping a single family issue open for both variants.

Next steps

• Investigate why healthcheck can exceed 60s under CI load in NoAgent scenario and consider:
  • Further increasing/relaxing the healthcheck wait in tests or mocking health endpoint for this path.
  • Ensuring health is cached or fetched with a shorter code path for the support bundle.

[johnstcn]

Fixed by coder/coder#16174

[ethanndickson]

Spotted again in a recent branch off main:
https://github.com/coder/coder/actions/runs/13621831271/job/38072525033

     support_test.go:150: 
        	Error Trace:	/Users/runner/work/coder/coder/cli/support_test.go:240
        	            				/Users/runner/work/coder/coder/cli/support_test.go:150
        	Error:      	Should NOT be empty, but was {0001-01-01 00:00:00 +0000 UTC false  {{<nil>  [] false} false map[] <nil> <nil> []} {{<nil>  [] false} false  false 0 } {false {<nil>  [] false}  0} {{<nil>  [] false} false false  0 0} {false {<nil>  [] false} {[]}} {{<nil>  [] false} []} }
        	Test:       	TestSupportBundle/NoAgent
        	Messages:   	health report should not be empty

[flake-investigator]

Reoccurrence details (2025-11-17)

• Workflow: Nightly gauntlet
• Run: https://github.com/coder/coder/actions/runs/19418075437
• Job: test-go-pg (macos-latest)
• Commit: 897286f33520f45e92f37c452c987124231319f2 (Ethan)

Evidence

=== FAIL: cli TestSupportBundle/NoAgent (63.30s)
... GET /api/v2/debug/health ... took=1m0.00276575s  status_code=503
response_body={"message":"Healthcheck is in progress and did not complete in time. Try again in a few seconds."}
... support_test.go:243 ... support_test.go:153
Error: Should NOT be empty, but was {...}
Test: TestSupportBundle/NoAgent
Messages: health report should not be empty

Classification: Flaky test (health check timeout). Assigning to @johnstcn as primary maintainer of support bundle tests.

Not infra; other packages passed, and the same pattern was fixed before by increasing healthcheck timeout (a8fbe71). Suggest revisiting timeout/mocking for NoAgent variant.