Repository navigation
Conversation
…emory with other processes) The install security check (#508) refuses ipc: host without privileged consent, and the read-only prod dry run of its Start check flagged this one bundle. llama-server is a single process (--no-mmap, weights in GPU memory through Vulkan) and uses no POSIX shared memory with other processes, so the host IPC namespace is not needed; the private /dev/shm keeps shm_size 16g as headroom. Manifest gets a version (1.0.1). Claude-Session: https://claude.ai/code/session_01N9fXzvNy8rCEArj1y5p6fp
kh0pper
force-pushed
the
fix/35b-shm-not-ipc-host
branch
from
October 21, 2026 01:58
3a589a9 to
25d8f3e
Compare
kh0pper
added a commit
that referenced
this pull request
Oct 21, 2026
…ass the check; operatorRoots now count (#524) Stacked on #509, so the test's ipc: host exception for the 35B is dead and is removed: all model bundles in the repo must pass the install security check as the orchestrator runs it. The preflight comment said the check ignores ctx.operatorRoots; since #524 the host-path allowlist uses them (a mount outside Crow's home must sit under one, and the floor applies). Claude-Session: https://claude.ai/code/session_01N9fXzvNy8rCEArj1y5p6fp
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This is a follow-up to #508. A read-only dry run of #508's Start check on prod passed every bundle except
llamacpp-vulkan-qwen36-35b-a3b, because it usesipc: hostwithout being privileged.Change: remove
ipc: hostfrom the 35B compose file and keepshm_size: 16g. Underipc: hostthat setting was ignored; it now sizes the container's own/dev/shm. The manifest also gets a version (1.0.1).Why
ipc: hostisn't needed: llama-server is one process. It loads weights with--no-mmapinto GPU memory through Vulkan, and it shares no POSIX shared memory or SysV IPC with any other process.ipc: hostis a common default in GPU containers for multi-process PyTorch data loaders. It does nothing for llama.cpp, and the private 16 GB/dev/shmis generous headroom.⚠ Verify in the next model window before merging.
servers/gateway/gpu-orchestrator.jsstarts this container from the repo's compose file. Once this is merged and pulled, the next on-demandcompose upsees a config change and recreates the container.Not in this PR:
llamacpp-vulkan-qwen3-embedandvllm-rocm-qwen35-4balso useipc: host. Both are started only by gpu-orchestrator, whose compose path is unguarded; that follow-up is recorded in queue item 30. vLLM may genuinely want a large shm (shm_size) instead.Tests: a new test in
tests/compose-guard.test.js, plus the existing tests that reference the 35B bundle. Registry rebuilt;check-port-allocationpasses.https://claude.ai/code/session_01N9fXzvNy8rCEArj1y5p6fp