row w1-r-ls-iq3 (R-LS-IQ3, ROW_KIND=llama-server) · started 2026-10-06T07:38:04Z · uid 1000 · python 3.14.4
CUDA_DEVICE_ORDER=PCI_BUS_ID CUDA_VISIBLE_DEVICES=0 TZ=UTC
pins: 34 files read their sha256 again in 0.5 s (the 2 shards were read at the window's open)
identity before the launch: the client's nine prompt digests equal L-IQ3-t8's (9/9)
port withheld-2026-10-13 reads free on the box (ss -ltn)
server argv: /workshop/bench-flashnext/llama.cpp-3cf03257/build/bin/llama-server -m /workshop/bench-flashnext/models/IQ3_S/Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf --host 127.0.0.1 --port withheld-2026-10-13 -ngl 99 --n-cpu-moe 0 -c 131072 -fa on -ctk q8_0 -ctv q8_0 -lm mmap --lazy-mode on -fit off -np 1 --cache-ram 0 --no-webui --jinja -lv 4 -t 8 -tb 8
the second fence (the unit's IP rules), inner.py's ipfilter probe: enforced (a connect to the box's own non-loopback address is dropped (the IP filter is enforced); loopback answers)
sandbox: /usr/bin/bwrap --unshare-net --unshare-pid --die-with-parent --ro-bind / / --dev-bind /dev /dev --proc /proc --tmpfs /run --tmpfs /tmp --bind /workshop/desktop-window-1/row/R-LS-IQ3.run /workshop/desktop-window-1/row/R-LS-IQ3.run --ro-bind /run/nvidia-persistenced /run/nvidia-persistenced -- /usr/bin/python3 /workshop/desktop-window-1/client/inner.py --spec /workshop/desktop-window-1/row/R-LS-IQ3.run/spec.json
sandbox note: /var/run is a link to /run (or absent): the /run tmpfs covers it
sandbox note: /run/nvidia-persistenced bound read-only (the driver's persistence socket)
cards before:
index, pci.bus_id, name, memory.used [MiB], memory.total [MiB], enforced.power.limit [W], temperature.gpu
0, withheld-2026-10-13, NVIDIA RTX PRO 6000 Blackwell Workstation Edition, 2 MiB, 97887 MiB, 300.00 W, 35
1, withheld-2026-10-13, NVIDIA GeForce RTX 3090 Ti, 18097 MiB, 24564 MiB, 350.00 W, 39
placement: whole -- every layer on the card (CUDA0); the input lookup tables in host memory: token_embd 322.07 MiB and the n-gram table 27465.95 MiB, as registered
server maps: /workshop/bench-flashnext/cuda-13.4/targets/x86_64-linux/lib/libcublas.so.13.8.0.4, /workshop/bench-flashnext/cuda-13.4/targets/x86_64-linux/lib/libcublasLt.so.13.8.0.4, /workshop/bench-flashnext/cuda-13.4/targets/x86_64-linux/lib/libcudart.so.13.4.92, /workshop/bench-flashnext/llama.cpp-3cf03257/build/bin/libggml-base.so.0.24.0, /workshop/bench-flashnext/llama.cpp-3cf03257/build/bin/libggml-cpu.so.0.24.0, /workshop/bench-flashnext/llama.cpp-3cf03257/build/bin/libggml-cuda.so.0.24.0, /workshop/bench-flashnext/llama.cpp-3cf03257/build/bin/libggml.so.0.24.0, /workshop/bench-flashnext/llama.cpp-3cf03257/build/bin/libllama-common.so.0.4.1, /workshop/bench-flashnext/llama.cpp-3cf03257/build/bin/libllama-server-impl.so, /workshop/bench-flashnext/llama.cpp-3cf03257/build/bin/libllama.so.0.4.1, /workshop/bench-flashnext/llama.cpp-3cf03257/build/bin/libmtmd.so.0.4.1, /usr/lib/x86_64-linux-gnu/libcuda.so.595.84
cards after:
index, pci.bus_id, name, memory.used [MiB], memory.total [MiB], enforced.power.limit [W], temperature.gpu
0, withheld-2026-10-13, NVIDIA RTX PRO 6000 Blackwell Workstation Edition, 2 MiB, 97887 MiB, 300.00 W, 62
1, withheld-2026-10-13, NVIDIA GeForce RTX 3090 Ti, 18097 MiB, 24564 MiB, 350.00 W, 42
ended 2026-10-06T07:45:03Z · inner rc 0 · last phase done · identity 9/9 · placement whole · fences held ['netns', 'unit_ip_filter']