Release LoadGen Api, Appliance and MCP Server: 1.0.0.12157
Release LoadGen Api, Appliance and MCP Server: 1.0.0.12157
LoadGen General
No general changes in this release.
LoadGen Api
Agent Fleet — readiness and assurance for your agents and appliances. A new Agent Fleet capability continuously gauges whether your fleet can carry load: it samples host utilization, runs a scheduled synthetic browser canary, and reports whether the fleet has capacity to spare or is saturated, aggregating a Fleet Readiness view backed by reusable Appliance Profiles and a Linux preflight check. Managed from the frontend.
[#9235] [#9236] Scheduling fixes for load-profile schedules. An Enabled schedule now computes its Next Run and actually triggers (previously "Next Run" stayed "-" and the schedule never ran), and schedules saved before the fix self-heal on the first tick after upgrade with no re-save. The wizard's schedule Timezone now persists instead of reverting to the placeholder, and next-run times honour the chosen timezone.
[#9245] Fixed classic-technology load profiles (RDS, WVD, VMware, Citrix) showing every assigned user twice — one row with a workload and a duplicate "-- none --" row — in the Users & Agents step. Duplicate bindings are merged (a workload-bearing assignment wins), and already-corrupted profiles heal on the next save.
[#9223]Credential Profiles — admin reveal, and lockdown for platform-issued tokens. An administrator with the credential-profiles admin grant can now reveal a platform-issued token mirror (for example to re-read a reusable enrollment token). At the same time these mirrors are tagged platform-issued, hidden from non-admin readers, blocked from edit/delete, and can no longer be referenced by a test or datasource definition to read back a live key.
RBAC — new endpoints are visible after upgrade. Built-in role grants now reconcile at startup, so a role no longer hides endpoints added in a new release until it is manually refreshed.
[#9187] Fixed Full, Core and VDI agents intermittently showing Offline with "did not respond within 5s" (a regression), and reachable-but-slow agents being evicted and left stuck Offline. Status now waits for the readiness handshake and no longer poisons the connection, with a clearer reason when an agent really is unreachable.
Platform stability under load. Fixed platform stalls and "request was canceled" errors under heavy SessionSight ingest; SessionSight ingestion is decoupled from the request thread pool so it no longer starves API request handling.
LoadGen Appliance
[#9209] Fixed reporting failing on every Linux appliance — the reporting worker required a Windows-only runtime and restart-looped at launch, so report generation never ran. (Distinct from the earlier Let's Encrypt HTTPS-wizard fix tracked under the same ticket.) Requires an API redeploy.
[#9222] Fixed the "License Required" page reappearing on an always-on appliance that holds a valid license. The appliance now refreshes its license from the cloud proactively before its offline window lapses, instead of only once a month, so the licensed state no longer flaps mid-month. Air-gapped and offline-activated modes are unchanged.
Update All — no more false failure or blind progress. Fixed Update All reporting a false "failed" when a leftover reconcile from phase 1 raced phase 2, and the progress modal sitting blind at 0% during the agent restart. Requires an Appliance Agent redeploy.
Accurate idle CPU. Fixed the appliance detail page reporting a high CPU figure (around 97%) on an idle box, where it was measuring its own monitoring probe rather than steady-state load.
LoadGen Core Agent Appliance
[#9175] Getting a Core Agent Appliance. The Lightweight Core Agent Appliance — a virtual appliance that adds dedicated load-generation capacity to your Platform — is now available. Request the image for your hypervisor (Hyper-V, VMware ESXi, VirtualBox or Azure) from LoadGen Support using the new request option in the frontend, deploy it, and pair it with your Platform using an enrollment token and administrator approval. Appliances need no inbound internet access, and all updates are pushed to them centrally from the Platform.
[#9234] Fixed a load test failing with "No Core Agent connected back to the results listener (port 4851) within 90s" even though the appliance showed Online. On an appliance that also runs Docker or a Windows guest, an unroutable virtual-network address could be advertised as the callback IP; the appliance now selects the interface with a real default route so the Core Agent can dial back.
[#9241] Fixed Manage → Static IP giving no success confirmation and silently auto-reverting after about 60 seconds. The dialog can now recover and confirm an in-flight IP change (even after the address has moved), instead of the confirm prompt appearing only once the rollback window had already elapsed.
Console improvements (#9225, #9228, #9229, #9230, #9240). The appliance console is easier to drive: a first-run setup order (enrollment token → Platform URL → register → test) with inline guidance, a Set timezone option, a Manage local hosts-file entries option (parity with the main appliance), a Logout option that closes the SSH session in one step, and a working cancel in the Set-static-IP flow (leave any prompt blank to abort).
[#9231] Fixed installing Google Chrome or Microsoft Edge (and the Playwright Firefox/WebKit dependencies) failing with a permissions error on a hardened appliance; browser installs now run so that the package manager keeps its own sandbox intact and succeed. (Covers the duplicate Edge report [#9232].)
[#9242] Browser installs now work on an appliance that reaches the internet only through an outbound proxy (the reachability check is proxy-aware), and, as a fallback, a browser can be installed offline by uploading the vendor package through the platform.
[#9243] Added the ability to permanently delete a revoked enrollment token (a still-valid token is refused).
Reliability fixes. A Core Agent Appliance no longer goes silently Offline after an agent self-update that previously wiped its platform registration; switching from DHCP to a static IP no longer drops DNS; and start/stop/restart actions for the Core Agent and Windows guest now report a real success or failure instead of always returning an empty success.
Management parity with the main appliance. The appliance's Manage view gained recent-jobs history, per-appliance outbound proxy configuration, host OS updates, disk and Docker status, host reboot, and local hosts-file management.
Windows and desktop testing on the appliance (new). A Core Agent Appliance can now host a Windows Full Agent guest, extending it beyond headless Web/Core to Windows-based and desktop-technology load testing — Citrix, RDS/WVD, VMware and FreeRDP — alongside its Core Agent capacity. The guest auto-registers as a Full Agent with the platform.
Windows cockpit. A new Windows cockpit in the appliance's Manage view manages the guest end to end: a dual-agent overview; version-picked install of the Full Agent, Insight and FreeRDP; Citrix Workspace management with version pinning; guest inventory, dual logs and diagnostic log levels; features & firewall; hostname and timezone; a local hosts-file editor; and installation of custom trusted root certificates.
Windows guest templates. Save a guest configuration once and apply it to many appliances, including KMS activation and an appliance-independent software catalog.