Intel GPUs (intel_gpu_top) never report temperature or memory, so the
"suspended card" heuristic in calculateGPUAverage (temp == 0 &&
memoryUsed == 0) fired on every collection that landed between samples.
intel_gpu_top samples every 3.3s (intelGpuStatsInterval) while the hub's
realtime worker collects every 1s, so most realtime collections had no
new sample (delta count 0) and returned an empty GPUData with power
omitted (json "p"/"pp" are omitempty). The frontend derives the GPU
Power Draw series and legend from the latest sample, so the chart and
legend blanked on roughly two of every three or four one-second cycles.
NVIDIA/AMD were unaffected because they report temperature even when
idle, so the heuristic never fired and the last average was already
carried forward.
Gate the zero-return on non-engine (discrete) GPUs so Intel GPUs carry
the last average forward during between-sample gaps, matching the
existing NVIDIA/AMD behavior. Add a regression test.
intel_gpu_top does not support the xe driver, so skip it for xe devices
and let the existing nvtop last-resort collector handle them. nvtop leaves
device_name unset on xe, so name the GPU from its PCI device id ("Intel GPU
(<id>)"). Adds nvtop to the Intel agent image (gputop already ships with
igt-gpu-tools).
- Prefix Intel GPU ID as i0 to avoid NVML/NVIDIA index IDs like 0
- Update frontend GPU engines chart to select a GPU by id instead of
assuming g[0]
- Adjust tests to use the new Intel GPU id
purego requires dynamic linking, so split the agent builds:
- Default: static binary without NVML (works on musl/alpine)
- Glibc: dynamic binary with NVML support via purego
Changes:
- Add glibc build tag to conditionally include NVML code
- Add beszel-agent-linux-amd64-glibc build/archive in goreleaser
- Update ghupdate to use glibc binary on glibc systems
- Switch nvidia dockerfile to golang:bookworm with -tags glibc
- add one minute charts
- update disk io to use bytes
- update hub and agent connection interfaces / handlers to be more
flexible
- change agent cache to use cache time instead of session id
- refactor collection of metrics which require deltas to track
separately per cache time
Luke WassandGitHub
Miłosz KolberandGitHub
T.J. TarazevitsandGitHub
henrygd
Robert AccetturaandGitHub
Vascolas007andGitHub
crimistandGitHub