Fix rosa deprovision

[rhopp]

when rosa provisioning task times out, the task result (ocp login) is empty - preventing the finalize task to deprovision the cluster. That leaves orphan clusters and once they accumulate, it prevents starting new ones.

[konflux-ci-qe-bot]

@rhopp: The following matrix E2E test has Failed, say /retest to rerun failed tests.

Total child pipelines: 2
Failed child pipelines: 2

Failed child pipelines

PipelineRun Name	Status	Rerun command	Build Log	Test Log
tsf-e2e-4.20-cv2t7	Failed	/retest	View Pipeline Log	View Test Logs
tsf-e2e-4.21-twv72	Failed	/retest	View Pipeline Log	View Test Logs

[rhopp]

/retest

[konflux-ci-qe-bot]

@rhopp: The following matrix E2E test has Failed, say /retest to rerun failed tests.

Total child pipelines: 2
Failed child pipelines: 2

Failed child pipelines

PipelineRun Name	Status	Rerun command	Build Log	Test Log
tsf-e2e-4.20-ndz69	Failed	/retest	View Pipeline Log	View Test Logs
tsf-e2e-4.21-gpgf7	Failed	/retest	View Pipeline Log	View Test Logs

[konflux-ci-qe-bot]

@rhopp: The following matrix E2E test has Failed, say /retest to rerun failed tests.

Total child pipelines: 2
Failed child pipelines: 2

Failed child pipelines

PipelineRun Name	Status	Rerun command	Build Log	Test Log
tsf-e2e-4.20-smc7g	Failed	/retest	View Pipeline Log	View Test Logs
tsf-e2e-4.21-ksxzd	Failed	/retest	View Pipeline Log	View Test Logs

[dperaza4dustbit]

I would avoid aws configure, this writes to file system. I think using env vars in memory should be enough: https://docs.aws.amazon.com/cli/v1/userguide/cli-configure-envvars.html

[Roming22]

/lgtm

[Roming22]

@rhopp Looks fine to me, but the CI is failing.

[rhopp]

/retest

[konflux-ci-qe-bot]

@rhopp: The following matrix E2E test has Failed, say /retest to rerun failed tests.

Total child pipelines: 2
Failed child pipelines: 2

Failed child pipelines

PipelineRun Name	Status	Rerun command	Build Log	Test Log
tsf-e2e-4.20-gsp5z	Failed	/retest	View Pipeline Log	View Test Logs
tsf-e2e-4.21-qvmdf	Failed	/retest	View Pipeline Log	View Test Logs

[xinredhat]

I think we just need one-line fix in rosa-hcp-provision task.

The fix: add a fallback at the end of the provision step script, after the main function call. Since the step uses onError: continue, Tekton will continue even if main fails — so code after it will execute.

Current (end of the provision step script):

          main 2>&1 | tee cluster-provision.log

Proposed (add fallback after the main call):

          main 2>&1 | tee cluster-provision.log

          # Ensure ocp-login-command is always set so finally tasks can resolve it
          if [ ! -s "$(results.ocp-login-command.path)" ]; then
            echo "echo 'No login command available - cluster provisioning failed'" > "$(results.ocp-login-command.path)"
          fi

[rhopp]

@xinredhat That's indeed much simpler, but I don't think it will work in the timeout scenario. When the taskrun runs for more than an hour, tekton just "kills" it, so no "dummy" result will get written.