Files
llmgw/internal/agent/engines/sglang.go
T
ilkerandClaude Opus 5 6a13e3dddf Phase 2 (1/2): node discovery, engine adapters, process supervisor
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>
2026-09-16 02:01:59 +02:00

69 lines
2.0 KiB
Go

package engines
import "strconv"
// SGLangConfig locates the SGLang installation.
type SGLangConfig struct {
// Python is the interpreter of the SGLang virtualenv. SGLang is launched as
// a module rather than a console script, which is the documented entry
// point and avoids depending on the script being on PATH.
Python string `koanf:"python"`
}
// SGLang builds SGLang launch commands.
type SGLang struct{ cfg SGLangConfig }
// Name identifies the engine in the catalog.
func (*SGLang) Name() string { return "sglang" }
// HealthPath is SGLang's readiness endpoint.
func (*SGLang) HealthPath() string { return "/health" }
// MetricsPath is SGLang's Prometheus endpoint.
func (*SGLang) MetricsPath() string { return "/metrics" }
// Validate checks the profile against the GPUs it would run on.
func (*SGLang) Validate(p Profile, gpus []GPUInfo) error {
if err := validateCommon(p, gpus); err != nil {
return err
}
if q := quantizationOf(p.Args); q == "fp8" {
if err := requireComputeCapability(p, gpus, 8.9, "fp8 quantization"); err != nil {
return err
}
}
return nil
}
// BuildCommand produces the launch command for one shard.
//
// SGLang supports multi-node natively, so unlike vLLM there is no second
// process to sequence: every rank runs the same command with its own rank number.
func (g *SGLang) BuildCommand(s ShardSpec) (Command, error) {
python := g.cfg.Python
if python == "" {
python = "python3"
}
p := s.Profile
args := []string{"-m", "sglang.launch_server",
"--model-path", p.ModelPath,
"--host", "0.0.0.0",
"--port", strconv.Itoa(s.HTTPPort),
"--tp", strconv.Itoa(p.TP),
}
if p.ServedModelName != "" {
args = append(args, "--served-model-name", p.ServedModelName)
}
if s.NNodes > 1 {
args = append(args,
"--nnodes", strconv.Itoa(s.NNodes),
"--node-rank", strconv.Itoa(s.Rank),
"--dist-init-addr", s.LeaderAddr+":"+strconv.Itoa(s.LeaderPort),
)
}
args = append(args, p.Args...)
return Command{Path: python, Args: args, Env: materializeEnv(baseEnv(s))}, nil
}