ilkerandClaude Opus 5
6a13e3dddf
Everything here works without GPUs and is covered by tests. The control plane connection is not built yet, so llmgw agent itself is still a stub and the phase exit criterion is not met. - discovery: IB from /sys/class/infiniband, parsing the formats sysfs actually uses and telling native InfiniBand from RoCE; GPUs via nvidia-smi, the cgo-free path, with health flags for uncorrected ECC and memory already held by another process. A node without IB is not an error, it is a single-node island. - engines: vLLM and SGLang command and environment construction, kept free of process handling so it stays testable. The NCCL environment is the part the plan marks critical: NCCL_IB_HCA lists only the chosen fabric's adapters, because leaving it open lets NCCL pick an adapter on a fabric where the peer is unreachable, and the job then hangs instead of failing. - validation before launch: no mixed GPU types in a tensor-parallel group, the node/GPU/TP arithmetic has to add up, and fp8 is refused below compute capability 8.9 rather than silently falling back to a slower path. - supervisor: process groups, because engines fork workers that hold GPU memory and outlive a signal sent only to the leader; a bounded log tail, since engine output is the only diagnosis when loading fails; Starting separate from Loading, because a 70B model legitimately takes minutes; and orphan adoption through pid files fingerprinted with the command line, since pids get reused and adopting an unrelated process is worse than leaking the original. - readiness is two questions: /health says the process is up, /v1/models says the weights loaded. Routing on /health alone sends traffic into a loading model. - llmgw agent inventory reports what the scheduler would see on a node. One plan correction: GPU class sizes come from the driver in GiB, which does not always match the marketing name. An L40S is sold as 48GB and reports 44.99 GiB, so its class is l40s-45g, not l40s-48g. Consistency matters more than matching the box, because that number decides whether a model fits. Plan §2 and §5.1 updated. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
58 lines
1.5 KiB
Go
58 lines
1.5 KiB
Go
// Package supervisor runs and watches engine processes (plan §5.4).
|
|
package supervisor
|
|
|
|
import "fmt"
|
|
|
|
// State is where a shard is in its lifecycle.
|
|
//
|
|
// Loading is separate from Starting because model loading dominates startup:
|
|
// a 70B model takes minutes, and an operator watching a shard sit in one state
|
|
// for ten minutes needs to know whether that is normal.
|
|
type State int
|
|
|
|
const (
|
|
// Pending means the shard is known but not started.
|
|
Pending State = iota
|
|
// Starting means the process has been launched but has not opened its port.
|
|
Starting
|
|
// Loading means the port is open but the model is not ready.
|
|
Loading
|
|
// Ready means health checks pass and the model is served.
|
|
Ready
|
|
// Draining means no new work; existing requests finish.
|
|
Draining
|
|
// Stopped means the process exited as intended.
|
|
Stopped
|
|
// Failed means the process exited unexpectedly or never became ready.
|
|
Failed
|
|
)
|
|
|
|
func (s State) String() string {
|
|
switch s {
|
|
case Pending:
|
|
return "pending"
|
|
case Starting:
|
|
return "starting"
|
|
case Loading:
|
|
return "loading"
|
|
case Ready:
|
|
return "ready"
|
|
case Draining:
|
|
return "draining"
|
|
case Stopped:
|
|
return "stopped"
|
|
case Failed:
|
|
return "failed"
|
|
default:
|
|
return fmt.Sprintf("unknown(%d)", int(s))
|
|
}
|
|
}
|
|
|
|
// Terminal reports whether the shard has finished for good.
|
|
func (s State) Terminal() bool { return s == Stopped || s == Failed }
|
|
|
|
// Live reports whether a process should be running in this state.
|
|
func (s State) Live() bool {
|
|
return s == Starting || s == Loading || s == Ready || s == Draining
|
|
}
|