Release LoadGen Api, Appliance and MCP Server: 1.0.0.11831
LoadGen General
- No general changes in this release.
LoadGen Api
- [#8514] Fixed datasources disappearing after a docker container restart on a fresh appliance. The configuration loader now accepts both the format the appliance agent writes and the legacy format, so datasources are always read back correctly. A malformed datasource file now produces an error-level log line instead of being silently swallowed as a warning.
- [#8493] Fixed uptime checks rendering with a misleading "No datasource" badge when the linked datasource was actually configured but its writes were failing. The list now distinguishes between "no datasource configured" and "datasource configured but write is failing", so the operator sees the real state at a glance.
- [#8502] Fixed the per-check uptime metrics endpoint reporting phantom open incidents for brand-new checks whose first probe failed. Incident counts and MTTR/MTBF are now sourced from the canonical incident registry instead of being heuristically reconstructed from raw probe rows, eliminating the disagreement between the dashboard summary and the per-check view.
- [#8486] Fixed SessionSight Visitors page returning zero results for sub-day windows immediately after activity. The visitors query now includes the current day's activity directly, so "I just visited" is reflected on the dashboard right away.
- [#8489] Fixed SessionSight heatmap pages showing "No pages found" with the default minimum-threshold of 5, even when the worker was actively producing rows. Click and scroll counts are now aggregated correctly before the threshold is applied, restoring the intended "minimum aggregated clicks per point / minimum sessions per scroll bucket" semantics.
- [#8487] Uptime maintenance windows extended end-to-end: the
Monthlyrecurrence type withdaysOfMonthis now wired through the API, validation, persistence and Swagger documentation. Monthly windows clamp values larger than the current month's last day, so an entry of "31" fires on the final day of every month. - [#8500] Improved diagnostics for license-not-available-after-reboot reports. License file read failures, license selection failures, and the runtime license identity are now logged at error level with full context so the next bundle pinpoints the failing condition.
LoadGen Appliance
- [#8513] Fixed the appliance failing to monitor LoadBots on the local network with
Connection ... did not become ready, while raw TCP from the appliance host worked fine. A misconfigured network rule on the appliance was silently dropping the traffic; monitoring of remote LoadBots is restored on every supported port. Existing appliances are healed automatically by a startup migration; no SSH access is required. - [#8491] Fixed the "Update All" job reporting Completed within seconds while the appliance and containers were still on the previous build. The job now stays in a "Running - awaiting agent restart" state until the underlying upgrade actually applies, and the existing stale-orchestration sweep transitions silently failed updates to Failed - so the Jobs UI tells the operator the truth instead of falsely showing green. New
POST /appliance/update-all/{id}/canceladmin endpoint provides an idempotent escape hatch for stuck orchestrations. - [#8490] Fixed clicking Update repeatedly returning a 409 Conflict ("Another update-all orchestration is already in phase 'UpdatingAgent'") even when the actual upgrade had already completed. A new background sweeper reaps stale orchestrations, and the same sweep also runs on demand the next time an operator clicks Update - so there is no need to SSH in to clear a stuck state file.
- [#8485] Reworked the appliance pre-login banner so the URL line only ever shows a valid address, and the LoadGen ASCII art was restored. A working URL or no URL at all - never a value the browser will reject.
LoadGen MCP Server
- No MCP Server changes in this release.