Release notes
OpsProbe v0.3.28: fewer false alerts.
Scheduled checks now confirm transient failures and require consecutive results before changing incident state.
Summary
HTTP, DNS, and TCP transport failures receive one bounded retry. An incident opens after two consecutive down results and resolves after two consecutive up results.
Changes
- Failed scheduled results identify the HTTP, DNS, TLS, or TCP layer and retain the underlying error.
- HTTP response failures and certificate failures are not retried in the same run.
- An incident-opening SQL error was corrected.
Limits
This is a single-location confirmation policy, not multi-node voting. Alert latency may increase by one monitor interval. Redirect-chain diagnosis and content assertions remain planned.
Verification
- Rust formatting, tests, and release build
- JavaScript syntax and release checks
- No database migration or new configuration