イベントドリブンのサーバーレス コンピューティング プラットフォームを提供する Azure サービス。
Hi @Morimoto Ryosuke ,
Thank you for reaching out to Microsoft Q & A !
Your authenticated HTTP and SQL write/read tests, together with version-specific telemetry, show that the deployed version executed successfully for the requests tested. However, they do not establish that every instance is healthy or explain the ARM deployment failure.
For Recreate, Microsoft’s states that there is “No completion signal” and recommends monitoring instance logs to track when the original instances stop emitting logs. This is separate from the reported RuntimeFailed / ExpectationFailed / 04916 after 20 minutes—the documentation does not establish that this failure is expected or harmless.
I haven’t found Microsoft documentation that links this specific failure to FDM monitoring being skipped or the instance counters remaining 0. I would therefore focus the next validation on:
- Capture the complete ARM response before stopping the app. Adjust the observation script to handle
errorsas an array and preserve responses for the same deployment ID, including anyfailedInstancesLogs. Verify the UTC sequence of the failure and stop operation; the current evidence does not establish their ordering. - Repeat your authenticated HTTP and SQL checks during that same window, particularly after
RuntimeFailedappears while the app remains running. Correlate the release identifier andcloud_RoleInstancewith startup errors and exceptions. The 10 expected JSON 401 responses validate the rejection path, rather than the authenticated workflow. Instance-log checks can provide additional evidence, but the absence of errors alone does not prove that every instance is healthy. - Use the same validation for rollback, checking the previous release identifier, required configuration, and SQL compatibility. Microsoft’s explains that redeploying previous code does not undo database data or schema changes.
Until the discrepancy is explained, I would record the result as “application checks passed; deployment completion remains unverified”, rather than treating CLI exit code 0 as sufficient.
Could you share the redacted pre-stop ARM response, deployment ID, polling API version, FDM skip log entry, and UTC timeline? Also, do authenticated requests continue to succeed after RuntimeFailed appears? Please remove credentials, connection strings, and signed log URLs before posting.
Please "Upvote the Answer" if this information helped you. This will help us and others in the community as well.