reanimator/bee
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity
333 Commits 1 Branch 98 Tags
d1a6863ceb239cdbef323f0c8e825809503b7ff1
Commit Graph

11 Commits

Author SHA1 Message Date
Mikhail Chusavitin
7a843be6b0 Stabilize DCGM GPU discovery 2026-04-03 09:50:33 +03:00
Mikhail Chusavitin
4f9b6b3bcd Harden NVIDIA boot logging on live ISO 2026-03-31 11:37:21 +03:00
Mikhail Chusavitin
6dee8f3509 Add NVIDIA stress loader selection and DCGM 4 support 2026-03-31 11:15:15 +03:00
Michael Chus
5ee120158e fix(build): remove unused variant package lists before lb build 
live-build picks up ALL .list.chroot files in config/package-lists/.
After rsync, bee-nvidia.list.chroot, bee-amd.list.chroot, and
bee-nogpu.list.chroot all end up in BUILD_WORK_DIR — causing lb to
try installing packages from every variant (and leaving version
placeholders unsubstituted in the unused lists).

Fix: after copying bee-${BEE_GPU_VENDOR}.list.chroot → bee-gpu.list.chroot,
delete all other bee-{nvidia,amd,nogpu}.list.chroot from BUILD_WORK_DIR.

Also includes nomsi boot mode changes (bee-nvidia-load + grub.cfg).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-03-30 23:03:42 +03:00
Mikhail Chusavitin
bd94b6c792 fix(iso): add libnvidia-ptxjitcompiler + ldconfig for PTX JIT and NCCL 
- 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>
 2026-03-26 10:37:27 +03:00
Mikhail Chusavitin
30cf014d58 Rename NVIDIA bootloader modes 2026-03-25 19:12:26 +03:00
Mikhail Chusavitin
27d478aed6 Add bootloader choice for safe vs full NVIDIA boot 2026-03-25 19:11:15 +03:00
Mikhail Chusavitin
d36e8442a9 Stabilize live ISO consoles and NVIDIA boot path 2026-03-25 19:05:18 +03:00
Mikhail Chusavitin
b6619d5ccc fix(iso): skip NVIDIA module load when no NVIDIA GPU present 
Check PCI vendor 10de before attempting insmod — avoids spurious
nvidia_uvm symbol errors on systems without NVIDIA hardware (e.g. AMD MI350).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-03-25 13:38:31 +03:00
Mikhail Chusavitin
b25a2f6d30 feat: add support bundle and raw audit export 2026-03-16 18:20:26 +03:00
Mikhail Chusavitin
345a93512a migrate ISO build from Alpine to Debian 12 (Bookworm) 
Replace the entire live CD build pipeline:
- Alpine SDK + mkimage + genapkovl → Debian live-build (lb config/build)
- OpenRC init scripts → systemd service units
- dropbear → openssh-server (native to Debian live)
- udhcpc → dhclient for DHCP
- apk → apt-get in setup-builder.sh and build-nvidia-module.sh
- Add auto/config (lb config options) and auto/build wrapper
- Add config/package-lists/bee.list.chroot replacing Alpine apks
- Add config/hooks/normal/9000-bee-setup.hook.chroot to enable services
- Add bee-nvidia-load and bee-sshsetup helper scripts
- Keep NVIDIA pre-compile pipeline (Option B): compile on builder VM against
  pinned Debian kernel headers (DEBIAN_KERNEL_ABI), inject .ko into includes.chroot
- Fixes: native glibc (no gcompat shims), proper udev, writable /lib/modules,
  no Alpine modloop read-only constraint, no stale apk cache issues

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
 2026-03-08 18:01:38 +03:00