Release v1.6.0

[rdwr-egore]

Container Alert v1.6.0 release candidate (since v1.5.4)

Status: Open release candidate; not merged, tagged, published, or deployed. The VERSION file remains 1.6.0. Full inventory: v1.6.0 release notes.

Included changes

• PR Beta #3 — reliability and installer hardening, concurrent state save, bounded manual exec probes, post-reboot summary.

• PR fix: Rate-limit removed container event warnings and resolve by ID #6 — event-ID resolution, name reuse, rate-limited removed-container warnings.

• PR fix: Deduplicate restart-loop alerts and require stable recovery #5 — authoritative Docker RestartCount, restart-loop deduplication, stable recovery.

• PR fix: Correlate planned Docker stops before reporting crashes #7 — intentional stop correlation, SIGKILL exit-137 vs confirmed OOM.

• PR feat: Optional HA active-node change alert (draft) #4 — optional independent HA active-node detection using promoted HA Operator.

• PR docs: clarify unreleased changes and image versioning #8 — release documentation/version clarification.

• PR Fix notification consistency and Syslog alert priorities for v1.6 #10 — notification field/priority consistency and Syslog retention at local WARNING/ERROR.

• Release branch — stable recovery for crashes/OOM even after external restarts, while preserving the incident state machine and HA logic.

• Customer communication — approved messages for CRASH/OOM/UNHEALTHY/RESTART-LOOP/RECOVERED/REBOOT-SUMMARY and fallback; retain HA-FAILOVER advice unchanged. Email/Slack human-facing label is Context; compatibility-preserving Syslog action=, event IDs, MIB and severity mapping unchanged. CRASH retains the read-only docker logs diagnostic; no instructions to modify Compose, resource limits, or other production settings.

• Critical test safety — removed live/shared-engine destructive testing path and unsafe host-based README exercises. Release smoke runs exclusively on an ephemeral GitHub-hosted runner, with an empty, uniquely labelled, localhost-only disposable Docker Engine; tests must fail closed otherwise. All cleanup verifies test labels and leftovers. No operational Docker workloads are touched by this workflow.

• Single installer — every operator receives the same complete versioned package (installer + prebuilt image + checksum). Run bash install.sh; no Internet, Git, source build or online/offline choice needed on the target. Existing settings stay in place; only the watchdog is recreated.

• Release packaging — after approval and tagging on main, hosted CI builds the versioned image and publishes exactly two assets: container-alert-v.tar.gz and SHA256SUMS. Assets are uploaded to a draft, checked, and only then published.

• Documentation audit and navigation — one customer Installation and Operations guide + versioned release notes in docs/customer; developer/release workflows, HA design, historical QA and evidence isolated under docs/engineering. The old duplicate DEPLOYMENT_ADVANCED file was removed after preserving its SNMP integration and technical content. Release package includes customer documentation only.

Runtime acceptance — commit 136ce0a (October 15, 2026)

• Normal PR CI: 119/119 Python regressions passed, source/shell/Compose validation and Docker image build passed.
• Isolated real Docker acceptance: 15/15 checks passed with actual Docker events and kernel OOM, including repeated restarts, incident lifecycle, planned 137 stop, manual restart and health checks.
• Additional real OOM → external start → stable RECOVERED, after 120.1 seconds.
• 15/15 structured notifications delivered to the disposable UDP Syslog receiver.
• Smoke cleanup checks completed; no leftover test-labelled containers or volumes. Hosted workflow also removed its disposable Engine.
• Workflow machine-readable artifacts retained: two PASS reports, actual Docker event logs, Syslog packets and observer state.
• Earlier real Pacemaker lab HA switchover and failback, node A → node B → node A, correctly generated one WARNING on the newly active node. This is historical lab evidence from an earlier v1.6 candidate; no new HA operation was performed for this text/safety update.

Documentation/privacy update: Commit 7077dfe removes lab-specific hostnames, addresses, and identifiers from the tracked release documentation, JSON evidence, configuration examples, and tests. It also simplifies customer-facing verification. The privacy cleanup was validated at that commit. No runtime monitoring logic was changed.

Current release validation — commit 3ae5a0e (October 16, 2026)

• 10 fake-Docker installer tests, 4 single-package tests and 5 smoke-safety tests passed without touching live Docker.
• 98 internal documentation links, YAML/shell checks and package-version/content checks passed.
• GitHub CI passed: 133/133 tests and Docker build/Compose validation.
• Hosted isolated Docker acceptance is running at this update.

Complete documentation audit — commit 43b0d44 (October 16, 2026)

• Audited 11 Markdown documents and four historical JSON evidence files against live source/configuration and installer/release workflow.
• Customer: Installation and Operations, v1.6.0 release notes. Those are the only Markdown guides packaged for customers alongside README.
• Engineering: Development and Release, v1.6.0 checklist, HA design, hosted test procedure, historical acceptance and test evidence.
• Corrected stale PR feat: Optional HA active-node change alert (draft) #4/HA statuses, smoke workflow trigger description, release-note links in GitHub Releases, template endpoints and obsolete image-download comments. Preserved SNMPv3/MIB integration details and historical probe snapshot.
• 4 documentation tests, 5 package tests, 10 fake-Docker installer tests, 5 isolation safety tests locally passed. Verified the release archive contains no engineering reports or application source, with customer links intact. Only documentation, examples/comments, package selection and non-destructive tests changed.
• GitHub CI passed 138/138 automated tests, shell/Compose validation and image build at the new commit.
• Hosted isolated Docker smoke still executing at this update; it must pass before merging.

Release gates and limitations

• Obtain final code/release review approval and merge PR Release v1.6.0 #9 to main.
• After approval for public distribution, push v1.6.0 from main and verify that the hosted release workflow publishes the complete versioned archive and its checksum. PR merging alone does not publish.
• Deploy or upgrade lab/customer systems only under approved operations and verify rollback. No deployment is performed by this PR.

End-to-end SMTP, Slack and SNMP delivery to real customer destinations were not all verified; the earlier user-requested SMTP timeout exception remains. No destructive cluster host reboot, STONITH or network-partition tests were performed. An external lab Docker restarter was observed historically, but its initiator was not identified.

Safety advisory: Earlier nested-engine testing on a shared host triggered real watchdog notifications. Destructive acceptance must run only on disposable hosted CI, never on shared or operational hosts.