A replica says what its engine computes on, and llama.cpp's word on the CPU refuses a GPU placement

llama.cpp's start-up output is read as a shard becomes ready: the devices it
used, its backend, and the layers it offloaded. That goes up with READY in a
new ShardStatus.compute, is kept on the shard, and is shown by the admin
API, ctl replicas (ON) and the console (COMPUTE). A shard given GPUs whose
engine says it found no usable GPU or offloaded no layer is stopped and
failed, even where nvidia-smi cannot be asked.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Ilker ManapandClaude Opus 5.5 committed 2026-10-07 13:02:17 +02:00
1 parent f285aa0548
commit 0bd0d881c5
15 files changed
+517 -126

No files matched your search

+12 -5
View File
@@ -37,11 +37,12 @@ import (
type statusRecorder struct {
ch chan status
mu sync.Mutex
calls []adoptCall
adopted []reconciler.AdoptedShard
unclaim []string // shard ids to report back as unclaimed
fetches []fetchReport
mu sync.Mutex
calls []adoptCall
adopted []reconciler.AdoptedShard
unclaim []string // shard ids to report back as unclaimed
fetches []fetchReport
computes []reconciler.Compute
}
type status struct {
@@ -61,6 +62,12 @@ func (r *statusRecorder) OnShardStatus(replicaID, shardID string, st supervisor.
// fetches records download progress, so a test can assert that a fetch is
// visible from the control plane rather than only in the node's log.
func (r *statusRecorder) OnShardCompute(replicaID, shardID string, c reconciler.Compute) {
r.mu.Lock()
defer r.mu.Unlock()
r.computes = append(r.computes, c)
}
func (r *statusRecorder) OnFetchProgress(replicaID, shardID, source string, done, total int64) {
r.mu.Lock()
defer r.mu.Unlock()