clocks.max.graphics / clocks.max.memory CSV fields return exit status 2 on
RTX PRO 6000 Blackwell (driver 98.x), causing the entire gpu inventory query
to fail and clock lock to be skipped → normalization: partial.
Fix:
- Add minimal fallback query (index,uuid,name,pci.bus_id,vbios_version,
power.limit) that succeeds even without clock fields
- Add enrichGPUInfoWithMaxClocks: parses "Max Clocks" section of
nvidia-smi -q verbose output to fill MaxGraphicsClockMHz /
MaxMemoryClockMHz when CSV fields fail
- Move nvidia-smi -q execution before queryBenchmarkGPUInfo so its output
is available for clock enrichment immediately after
- Tests: cover enrichment and skip-if-populated cases
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- grub.cfg + isolinux/live.cfg.in: add pcie_aspm=off,
intel_idle.max_cstate=1 and processor.max_cstate=1 to all
non-failsafe boot entries
- bee-hpc-tuning: new script that sets all CPU cores to performance
governor via sysfs and logs THP state at boot
- bee-hpc-tuning.service: runs before bee-nvidia and bee-audit
- 9000-bee-setup.hook.chroot: enable service and mark script executable
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- collector/pcie: add applyPCIeLinkSpeedWarning that sets status=Warning
and ErrorDescription when current link speed is below maximum negotiated
speed (e.g. Gen1 running on a Gen5 slot)
- collector/pcie: add pcieLinkSpeedRank helper for Gen string comparison
- collector/pcie_filter_test: cover degraded and healthy link speed cases
- platform/techdump: collect lspci -vvv → lspci-vvv.txt for LnkCap/LnkSta
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Two issues:
1. BMC/management VGA chips (e.g. Huawei iBMC Hi171x, ASPEED) were included
in GPU inventory because shouldIncludePCIeDevice only checked the PCI class,
not the device name. Added a name-based filter for known BMC/management
patterns when the class is VGA/display/3d.
2. New NVIDIA GPUs (e.g. RTX PRO 6000 Blackwell, device ID 2bb5) showed as
"Device 2bb5" because lspci's database lags behind. Added "name" to the
nvidia-smi query and use it to override dev.Model during enrichment.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
hpl was listed in baseTargets and stressOnlyTargets but /api/sat/hpl/run
was never registered, causing a 405 Method Not Allowed (not valid JSON)
error when Validate one by one was triggered in stress mode.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add parallel GPU mode (checkbox, off by default): runs all selected GPUs
simultaneously via a single bee-gpu-burn invocation instead of sequentially;
per-GPU telemetry, throttle counters, TOPS, and scoring are preserved
- Make queryBenchmarkGPUInfo resilient: falls back to a base field set when
extended fields (attribute.multiprocessor_count, power.default_limit) cause
exit status 2, preventing lgc normalization from being silently skipped
- Log explicit "graphics clock lock skipped" note when inventory is unavailable
- Collect server model from DMI (/sys/class/dmi/id/product_name) and store in
result JSON; benchmark history columns now show "Server Model (N× GPU Model)"
grouped by server+GPU type rather than individual GPU index
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Tasks are now started simultaneously when multiple are enqueued (e.g.
Run All). The worker drains all pending tasks at once and launches each
in its own goroutine, waiting via WaitGroup. kmsg watcher updated to
use a shared event window with a reference counter across concurrent tasks.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Raise loglevel from 3 to 6 (INFO) and add systemd.show_status=1 so
kernel driver messages and systemd [ OK ]/[ FAILED ] lines are visible
during boot instead of showing only a blank cursor.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add platform/error_patterns.go: pluggable table of kernel log patterns
(NVIDIA/GPU, PCIe AER, storage I/O, MCE, EDAC) — extend by adding one struct
- Add app/component_status_db.go: persistent JSON store (component-status.json)
keyed by "pcie:BDF", "storage:dev", "cpu:all", "memory:all"; OK never
downgrades Warning or Critical
- Add webui/kmsg_watcher.go: goroutine reads /dev/kmsg during SAT tasks,
writes Warning to DB for matched hardware errors
- Fix task status: overall_status=FAILED in summary.txt now marks task failed
- Audit routine overlays component DB statuses into bee-audit.json on every read
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Support bundle is now built on-the-fly when the user clicks the button,
regardless of whether other tasks are running:
- GET /export/support.tar.gz builds the bundle synchronously and streams it
directly to the client; the temp archive is removed after serving
- Remove POST /api/export/bundle and handleAPIExportBundle — the task-queue
approach meant the bundle could only be downloaded after navigating away
and back, and was blocked entirely while a long SAT test was running
- UI: single "Download Support Bundle" button; fetch+blob gives a loading
state ("Building...") while the server collects logs, then triggers the
browser download with the correct filename from Content-Disposition
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
All final artefacts for a given version now land in one place:
dist/easy-bee-v4.1/
easy-bee-nvidia-v4.1-amd64.iso
easy-bee-nvidia-v4.1-amd64.logs.tar.gz ← log archive
(logs dir deleted after archiving)
- Introduce OUT_DIR="${DIST_DIR}/easy-bee-v${ISO_VERSION_EFFECTIVE}"
- Move LOG_DIR, LOG_ARCHIVE, and ISO_OUT into OUT_DIR
- cleanup_build_log: use dirname(LOG_DIR) as tar -C base so the path is
correct regardless of where OUT_DIR lives; delete LOG_DIR after archiving
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- tasks: mark TaskRunning tasks as TaskFailed on bee-web restart instead of
re-queueing them — prevents duplicate gpu-burn-worker spawns when bee-web
crashes mid-test (each restart was launching a new set of 8 workers on top
of still-alive orphans from the previous crash)
- server: reduce metrics collector interval 1s→5s, grow ring buffer to 360
samples (30 min); cuts nvidia-smi/ipmitool/sensors subprocess rate by 5×
- platform: add KillTestWorkers() — scans /proc and SIGKILLs bee-gpu-burn,
stress-ng, stressapptest, memtester without relying on pkill/killall
- webui: add "Kill Workers" button next to Cancel All; calls
POST /api/tasks/kill-workers which cancels the task queue then kills
orphaned OS-level processes; shows toast with killed count
- metricsdb: sort GPU indices and fan/temp names after map iteration to fix
non-deterministic sample reconstruction order (flaky test)
- server: fix chartYAxisNumber to use one decimal place for 1000–9999
(e.g. "1,7к" instead of "2к") so Y-axis ticks are distinguishable
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Burn tab: replace 6 flat cards with 3 grouped cards (GPU Stress,
Compute Stress, Platform Thermal Cycling) + global Burn Profile
- Run All button at top enqueues all enabled tests across all cards
- GPU Stress: tool checkboxes enabled/disabled via new /api/gpu/tools
endpoint based on driver status (/dev/nvidia0, /dev/kfd)
- Compute Stress: checkboxes for cpu/memory-stress/stressapptest
- Platform Thermal Cycling: component checkboxes (cpu/nvidia/amd)
with platform_components param wired through to PlatformStressOptions
- bee-gpu-burn: default size-mb changed from 64 to 0 (auto); script
now queries nvidia-smi memory.total per GPU and uses 95% of it
- platform_stress: removed hardcoded --size-mb 64; respects Components
field to selectively run CPU and/or GPU load goroutines
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
console=tty0 sent kernel messages to the active VT (tty1), overwriting
the TUI. Changed to console=tty2 so kernel logs land on a dedicated
console. tty1 is now clean; operator can press Alt+F2 to inspect kernel
messages and Alt+F3 for an extra shell.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- GPU Platform Stress Test now shows a live in-TUI chart instead of nvtop.
nvidia-smi is polled every second; up to 60 data points per GPU kept.
All three metrics (Usage %, Temp °C, Power W) drawn on a single plot,
each normalised to its own range and rendered in a different colour.
- Memory allocation changed from MemoryMB/16 to MemoryMB-512 (full VRAM
minus 512 MB driver overhead) so bee-gpu-stress actually stresses memory.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
loglevel=3 was hiding all kernel messages on tty0/ttyS0 except errors.
Machine crashes (panics, driver oops, module failures) were silent on VGA.
Restored loglevel=7 so kernel messages up to debug are printed to both
tty0 (VGA) and ttyS0 (SOL). Journald MaxLevelConsole reduced to info
(was debug) to reduce noise on SOL while keeping it useful.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Previously the ISO file was named after git describe --match 'audit/v*',
so a new iso/ tag produced names like v1.0.9-1-gXXXXXXX instead of v1.0.17.
Now build.sh has resolve_iso_version() that looks at iso/v* tags separately.
The bee binary inside the ISO still uses AUDIT_VERSION_EFFECTIVE.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- build-nvidia-module.sh: copy libnvidia-ptxjitcompiler.so.* alongside
libcuda/libnvidia-ml — required by cuModuleLoadDataEx for PTX JIT.
Without it: CUDA_ERROR_JIT_COMPILER_NOT_FOUND at runtime.
Cache check updated to force rebuild when ptxjitcompiler is missing.
- bee-nvidia-load: run ldconfig after module load so that NVIDIA/NCCL
libs injected into /usr/lib/ are visible to dlopen() callers.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
nvtop is no longer shown during NVIDIA SAT runs.
[o] Open nvtop shortcut also removed from the running screen.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Commit d36e844 dropped console=tty0 and added dual-serial + debug logging.
Without console=tty0 the kernel never initialises the VGA console,
leaving the physical screen permanently blank.
- Restore console=tty0 (VGA) as primary, keep console=ttyS0 for SOL
- Drop console=ttyS1 (redundant second serial port)
- Replace loglevel=7 + journald debug flood with loglevel=3 (errors only)
so kernel messages don't overwrite the TUI on the local screen
- Remove systemd.log_target/forward_to_console debug params
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
AMD does not publish Debian Bookworm packages at all (only focal/jammy/noble).
Switch ROCM_UBUNTU_DIST to "jammy"; jammy packages install cleanly on
Debian 12 due to compatible glibc. Also expand candidate list to include
point-releases (6.3.4, 6.3.3, …) so we pick the latest actually-published one.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Apply the same pattern as NVIDIA SAT: launch nvtop via tea.ExecProcess
so it occupies the full terminal as a live GPU chart (temp, power, fan,
utilisation lines) while the stress test runs in the background.
- Add screenGPUStressRunning screen + dedicated running/render handlers
- startGPUStressTest: tea.Batch(stress goroutine, tea.ExecProcess(nvtop))
- [o] reopen nvtop at any time; [a] abort (cancels context)
- Graceful degradation: test still runs if nvtop is not on PATH
- gpuStressDoneMsg routes result to screenOutput on completion
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
ROCm 6.4 does not yet publish a Release file for Debian Bookworm, causing
the live-build chroot hook to fail with "does not have a Release file".
Try each version in ROCM_CANDIDATES order; skip to the next if apt-get update
fails (repo unavailable). Exit gracefully if none are available.
Also rename inner 'candidate' variable to 'smi_path' to avoid collision.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Two-phase GPU thermal cycling test with per-second telemetry:
- Phases: baseline → load1 → pause (no cooldown) → load2 → cooldown
- Monitors: fan RPM (ipmitool sdr), CPU/server temps (ipmitool/sensors),
system power (ipmitool dcmi), GPU temp/power/usage/clock/throttle (nvidia-smi)
- Detects throttling via clocks_throttle_reasons.active bitmask
- Measures fan response lag from load start (validates case-04 ~2s lag)
- Exports metrics.csv (wide format, one row/sec) and fan-sensors.csv (long format)
- TUI: adds [F] Fan Stress Test to Health Check screen with Quick/Standard/Express modes
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- New tui/sat_progress.go: polls {DefaultSATBaseDir}/{prefix}-*/verbose.log every 300ms and parses completed/in-progress steps
- Busy screen now shows each step as PASS lscpu (234ms) / FAIL stress-ng (60.0s) / ... sensors-after instead of just "Working..."
2. Test results shown on screen (instead of just "Archive written to /path")
- RunCPUAcceptancePackResult, RunMemoryAcceptancePackResult, RunStorageAcceptancePackResult, RunAMDAcceptancePackResult now read summary.txt from the run directory and return a formatted per-step result:
Run: 2025-03-25T10:00:00Z
PASS lscpu
PASS sensors-before
FAIL stress-ng
PASS sensors-after
Overall: FAILED (ok=3 failed=1)
3. AMD GPU SAT with auto-detection
- platform.System.DetectGPUVendor(): checks /dev/nvidia0 → "nvidia", /dev/kfd → "amd"
- platform.System.RunAMDAcceptancePack(): runs rocm-smi, rocm-smi --showallinfo, dmidecode
- GPU SAT (G key / GPU row enter) automatically routes to AMD or NVIDIA based on detected vendor
- "Run All" also auto-detects vendor
4. Panel detail view
- GPU detail now shows the most recent (NVIDIA or AMD) SAT result, whichever is newer
- All SAT detail views use the same human-readable formatSATDetail format
- firmware-amd-graphics: Aldebaran firmware blobs (fixes amdgpu IB ring
test errors on MI250/MI250X at boot)
- 9001-amd-rocm.hook.chroot: adds AMD ROCm 6.4 apt repo and installs
rocm-smi-lib for GPU monitoring (analogous to nvidia-smi)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
--start is not a valid nvme-cli flag; correct syntax is -s 1 (short test).
Add --wait so the command blocks until the test completes.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
