Repository navigation
en US deployment verify
AKStream documentation edited this page Oct 7, 2026
·
1 revision
Deployment acceptance is executed following the end-user path, and verifiable evidence must be saved for every step. Monitoring service processes, health interfaces, or single APIs alone is insufficient to prove system availability.
flowchart LR
A[Version and services] --> B[Login and permission]
B --> C[Device onboarding and live view]
C --> D[Recording and playback]
D --> E[RTC / external integration]
E --> F[Security, monitoring, backup, rollback]
- The installation package version is consistent with the plan, and archive checksums have been verified.
- Versions of the program, WebUI, NodeAgent, and MediaServer have been recorded.
- Current node IDs, deployment roles, and the list of database and media nodes have been archived.
- No mixing of old configuration directories or uncleared test databases.
- systemd, Windows SCM, launchd, or Docker report that services are normal.
- NodeAgent is online, and the Provider is consistent with the target platform.
- After recovery verification in a maintenance window, the main process is restarted by the single recovery owner declared for the platform; Linux must prove NodeAgent recovery after the control connection is lost.
- Database, MediaServer, background tasks, and WebHooks recover after the main process restarts.
- Both liveness and readiness checks pass; Warnings have clear acceptance justifications, and blocking items have been cleared.
- The effective configuration version, disk configuration, and page display are consistent.
- Changes requiring a restart have been processed, and there are no long-pending restart items.
- Sensitive fields do not appear in logs, screenshots, or configuration diffs.
Select one real target model device:
- Complete registration or discovery, authentication, and channel synchronization.
- Start live preview and record the device, channel, media node, and stream identifiers.
- Verify HLS, FLV, RTSP, or WebRTC as required by the project.
- Verify stop and resource release.
- If required by the project, further verify PTZ, presets, image parameters, intercom, events, or GPS.
- Manual recording can start, stop, and produce files.
- Recording schedules trigger in the scheduled windows, and midnight-crossing rules execute as expected.
- Recording index can be queried; Range playback, download, and permissions are correct.
- Soft delete, recovery, and physical delete are verified only on test data.
- When clipping and merging are enabled, the output is playable, and temporary files can be cleaned.
- Alerts for disk capacity, inodes, and write permissions are configured.
- Verify separately from the real management network, device network, and end-user network.
- HTTPS certificate chains, domains, renewals, and WebSockets are normal.
- Public network RTC completes TURN or restricted network verification.
- Database ports such as 3306 are not open to unnecessary networks.
- Default passwords have been replaced; permissions are verified for administrator, operator, and read-only roles respectively.
- API Tokens use minimum privileges and can be revoked.
- Configurations,
Data/Security, databases, recording indexes, and authorization materials are included in the backup. - A recovery has been performed once in an isolated environment, recording the recovery time and gaps.
- Backup points before upgrades, old packages, and rollback commands are clearly defined.
- Monitoring can detect failures of the main service, Agent, MediaServer, database, and disk.
Record at least the following for each item:
- Test name, time, operator, and environment;
- Stable ID of the input object;
- Expected and actual results;
- Key screenshots;
- Command/Task ID, TraceId, and stream identifiers;
- Relevant log time windows;
- Whether it passed and any known limitations.
Sensitive passwords, Tokens, MediaServer secrets, authorization private keys, and full Cookies must not enter the evidence package.
All mandatory project paths must pass; items that did not pass must have a clear risk owner, impact scope, and resolution deadline; backup recovery and rollback must not be mere paper plans. After release, transition to Daily Inspection.