Stored ctl bench results reach the reconciler: requests per replica past the measured SLO concurrency scale a model up, a replica is taken back only if the rest stay under 70% of theirs, and among profiles of equal priority the most measured tokens per GPU or core is tried first. ctl model target sizes a model for a throughput against the room there is. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
267 lines
11 KiB
Go
267 lines
11 KiB
Go
package cli
|
|
|
|
import (
|
|
"encoding/json"
|
|
"fmt"
|
|
"os"
|
|
"os/user"
|
|
"time"
|
|
|
|
"github.com/knadh/koanf/parsers/yaml"
|
|
"github.com/knadh/koanf/providers/file"
|
|
"github.com/knadh/koanf/v2"
|
|
"github.com/spf13/cobra"
|
|
|
|
"gitea.mynodes.xyz/ilker/llmgw/internal/control/adminapi"
|
|
"gitea.mynodes.xyz/ilker/llmgw/internal/control/catalog"
|
|
"gitea.mynodes.xyz/ilker/llmgw/internal/sourcecheck"
|
|
)
|
|
|
|
// newModelCmd adds and removes models without a restart.
|
|
//
|
|
// Adding a model used to mean editing a file on the control plane and
|
|
// restarting it: a settling window, adoption churn, and on one occasion two
|
|
// minutes of outage. For a fixed production set that is a reasonable trade; for
|
|
// a platform whose promise is that people can try things, it is the difference
|
|
// between trying one and not bothering.
|
|
func newModelCmd(client func() *ctlClient, asJSON *bool) *cobra.Command {
|
|
c := &cobra.Command{
|
|
Use: "model",
|
|
Short: "Add or remove models while the cluster is running",
|
|
}
|
|
c.AddCommand(newModelAddCmd(client), newModelRemoveCmd(client), newModelPlanCmd(client),
|
|
newResourcesCmd(client), newClassCmd(client), newOwnerCmd(client),
|
|
newTargetCmd(client, asJSON),
|
|
newModelServeCmd(client))
|
|
return c
|
|
}
|
|
|
|
func newModelAddCmd(client func() *ctlClient) *cobra.Command {
|
|
var path string
|
|
var skipCheck bool
|
|
c := &cobra.Command{
|
|
Use: "add -f <file.yaml>",
|
|
Short: "Add a model from a definition file",
|
|
Long: "The definition is the same YAML a model takes in server.yaml, so an\n" +
|
|
"entry can be moved between the two by copying it.\n\n" +
|
|
"The catalog is rebuilt and validated before anything is written, so a\n" +
|
|
"definition that collides with an existing name, or names a fallback that\n" +
|
|
"does not exist, is refused rather than stored and then ignored on every\n" +
|
|
"start. Once accepted it takes effect on the reconciler's next pass — no\n" +
|
|
"restart — and survives one.",
|
|
Args: cobra.NoArgs,
|
|
RunE: func(cmd *cobra.Command, _ []string) error {
|
|
if path == "" {
|
|
return fmt.Errorf("--file is required: the model definition, as YAML")
|
|
}
|
|
// Read by koanf, the same parser and the same tags that read
|
|
// server.yaml — so an entry can be moved between the two by copying
|
|
// it, and a field that works in one cannot silently mean nothing in
|
|
// the other.
|
|
k := koanf.New(".")
|
|
if err := k.Load(file.Provider(path), yaml.Parser()); err != nil {
|
|
return fmt.Errorf("%s: %w", path, err)
|
|
}
|
|
var m catalog.Model
|
|
if err := k.UnmarshalWithConf("", &m, koanf.UnmarshalConf{Tag: "koanf"}); err != nil {
|
|
return fmt.Errorf("%s: %w", path, err)
|
|
}
|
|
if m.Name == "" {
|
|
return fmt.Errorf("%s: the definition has no name", path)
|
|
}
|
|
|
|
// Before the catalog is touched, because a source that does not
|
|
// resolve is otherwise discovered by a replica failing to start
|
|
// minutes later on a node — with an error naming the node rather
|
|
// than the typo.
|
|
if !skipCheck {
|
|
size, err := sourcecheck.Check(cmd.Context(), m.Source)
|
|
if err != nil {
|
|
return fmt.Errorf("%w\n Pass --no-source-check to add it anyway, "+
|
|
"which is right when the mirror is not up yet or the nodes "+
|
|
"reach it by an address this machine does not.", err)
|
|
}
|
|
if _, isURL := sourcecheck.HTTPSource(m.Source); isURL {
|
|
fmt.Fprintln(cmd.OutOrStdout(), sourceNote(m.Source, size))
|
|
}
|
|
}
|
|
|
|
body, err := json.Marshal(m)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
|
|
var res map[string]string
|
|
if err := client().post(cmd.Context(), "/api/v1/models", body, &res); err != nil {
|
|
return err
|
|
}
|
|
fmt.Fprintf(cmd.OutOrStdout(),
|
|
"%s added. It will be placed on the next reconcile pass; watch it with:\n"+
|
|
" llmgw ctl models\n", m.Name)
|
|
return nil
|
|
},
|
|
}
|
|
c.Flags().StringVarP(&path, "file", "f", "", "YAML file holding the model definition")
|
|
c.Flags().BoolVar(&skipCheck, "no-source-check", false,
|
|
"add it without asking whether the weights are actually at that URL")
|
|
return c
|
|
}
|
|
|
|
func newModelRemoveCmd(client func() *ctlClient) *cobra.Command {
|
|
return &cobra.Command{
|
|
Use: "remove <name>",
|
|
Short: "Remove a model that was added at runtime",
|
|
Long: "Only models added at runtime. One declared in server.yaml would come\n" +
|
|
"back on the next restart, so removing it here would leave the cluster\n" +
|
|
"disagreeing with its own configuration until then.\n\n" +
|
|
"Its replicas are stopped on the next pass, after the endpoint has been\n" +
|
|
"withdrawn, so traffic stops before the engine does.",
|
|
Args: cobra.ExactArgs(1),
|
|
RunE: func(cmd *cobra.Command, args []string) error {
|
|
var res map[string]string
|
|
if err := client().delete(cmd.Context(), "/api/v1/models/"+args[0], &res); err != nil {
|
|
return err
|
|
}
|
|
fmt.Fprintf(cmd.OutOrStdout(), "%s removed.\n", args[0])
|
|
return nil
|
|
},
|
|
}
|
|
}
|
|
|
|
// actor names who made a change, for the audit trail on a stored model.
|
|
func actor() string {
|
|
// Under sudo, the person who ran it rather than root.
|
|
if u := os.Getenv("SUDO_USER"); u != "" && os.Geteuid() == 0 {
|
|
return u
|
|
}
|
|
if u, err := user.Current(); err == nil && u.Username != "" {
|
|
return u.Username
|
|
}
|
|
return "unknown"
|
|
}
|
|
|
|
func newScaleCmd(client func() *ctlClient) *cobra.Command {
|
|
var min, max int
|
|
var idle time.Duration
|
|
c := &cobra.Command{
|
|
Use: "scale <model> --min N [--max M]",
|
|
Short: "Change how many replicas a model runs",
|
|
Long: "The replica count is --min. --max is a ceiling, not a target: a model\n" +
|
|
"with a minimum runs exactly that many, and nothing scales it up under\n" +
|
|
"load. Leaving --max out makes it follow --min.\n\n" +
|
|
"--min 0 with a --max above it is the on-demand model: nothing runs until\n" +
|
|
"somebody asks, and it stops again after the idle timeout. That loads one\n" +
|
|
"replica however high --max is. --min 0 --max 0 is a model kept in the\n" +
|
|
"catalog and deliberately not running.\n\n" +
|
|
"Only models added at runtime. One declared in server.yaml is read again\n" +
|
|
"on every start, so a number set here would hold until the next restart\n" +
|
|
"and then revert — change it in the file instead.\n\n" +
|
|
"It takes effect on the reconciler's next pass, and survives a restart.",
|
|
Args: cobra.ExactArgs(1),
|
|
RunE: func(cmd *cobra.Command, args []string) error {
|
|
if !cmd.Flags().Changed("min") {
|
|
return fmt.Errorf("--min is required: it is the replica count")
|
|
}
|
|
body, err := json.Marshal(adminapi.ScaleRequest{Min: min, Max: max, IdleTimeout: idle})
|
|
if err != nil {
|
|
return err
|
|
}
|
|
var res adminapi.ScaleResponse
|
|
if err := client().post(cmd.Context(),
|
|
"/api/v1/models/"+args[0]+"/scale", body, &res); err != nil {
|
|
return err
|
|
}
|
|
// What is in force, not what was asked for: the catalog fills in a
|
|
// max that was left out, and printing the request back would show a
|
|
// number the cluster is not using.
|
|
out := cmd.OutOrStdout()
|
|
switch {
|
|
case res.OnDemand:
|
|
fmt.Fprintf(out, "%s is on-demand: loaded when asked for, up to %d, stopped when idle.\n",
|
|
res.Model, res.Max)
|
|
case res.Min == 0:
|
|
fmt.Fprintf(out, "%s will not be placed; it stays in the catalog.\n", res.Model)
|
|
default:
|
|
fmt.Fprintf(out, "%s: %d replicas (max %d). It takes effect on the next pass; watch it with:\n"+
|
|
" llmgw ctl models\n", res.Model, res.Min, res.Max)
|
|
}
|
|
return nil
|
|
},
|
|
}
|
|
c.Flags().IntVar(&min, "min", 0, "replicas to keep running; this is the count")
|
|
c.Flags().IntVar(&max, "max", 0, "ceiling for an on-demand model; follows --min when left out")
|
|
c.Flags().DurationVar(&idle, "idle-timeout", 0,
|
|
"how long an on-demand model stays up unasked-for; zero takes the cluster default")
|
|
return c
|
|
}
|
|
|
|
func newResourcesCmd(client func() *ctlClient) *cobra.Command {
|
|
var memory, cores int
|
|
var profile string
|
|
c := &cobra.Command{
|
|
Use: "resources <model> --memory GB [--cores N]",
|
|
Short: "Set what a model reserves of a node: memory, and CPU cores",
|
|
Long: "--memory is the host RAM a replica needs: for a model with no GPU, its\n" +
|
|
"weights, KV cache and engine together. Placement reserves it on the node,\n" +
|
|
"so a node holds only as many replicas as it has memory for.\n\n" +
|
|
"--cores is the CPU cores a replica with no GPU computes on. They are\n" +
|
|
"reserved the same way, and become the engine's thread count, so two\n" +
|
|
"replicas on one node stop competing for the same cores. Zero reserves\n" +
|
|
"none and leaves the thread count to the node.\n\n" +
|
|
"Running replicas keep what they were started with; `llmgw ctl restart\n" +
|
|
"--model M` moves them to the new figures.",
|
|
Args: cobra.ExactArgs(1),
|
|
RunE: func(cmd *cobra.Command, args []string) error {
|
|
if !cmd.Flags().Changed("memory") {
|
|
return fmt.Errorf("--memory is required")
|
|
}
|
|
body, err := json.Marshal(adminapi.ResourcesRequest{Profile: profile, MemoryGB: memory, CPUCores: cores})
|
|
if err != nil {
|
|
return err
|
|
}
|
|
var m catalog.Model
|
|
if err := client().post(cmd.Context(), "/api/v1/models/"+args[0]+"/resources", body, &m); err != nil {
|
|
return err
|
|
}
|
|
for _, p := range m.Profiles {
|
|
fmt.Fprintf(cmd.OutOrStdout(), "%s/%s: %d GB memory, %d cores\n", m.Name, p.Name, p.MinMemoryGB, p.CPUCores)
|
|
}
|
|
fmt.Fprintf(cmd.OutOrStdout(), "Running replicas keep their old figures until restarted: llmgw ctl restart --model %s\n", m.Name)
|
|
return nil
|
|
},
|
|
}
|
|
c.Flags().IntVar(&memory, "memory", 0, "host RAM a replica needs, in GB")
|
|
c.Flags().IntVar(&cores, "cores", 0, "CPU cores a replica with no GPU computes on; 0 reserves none")
|
|
c.Flags().StringVar(&profile, "profile", "", "one profile; every profile when left out")
|
|
return c
|
|
}
|
|
|
|
func newClassCmd(client func() *ctlClient) *cobra.Command {
|
|
return &cobra.Command{
|
|
Use: "class <model> critical|standard|opportunistic|default",
|
|
Short: "Set how much a model's capacity is worth against another's",
|
|
Long: "When a model cannot be placed, replicas of a lower class are stopped to\n" +
|
|
"make room for it:\n\n" +
|
|
" critical never gives up the replicas of its minimum\n" +
|
|
" standard gives way to critical\n" +
|
|
" opportunistic gives way to both; the default for an on-demand model\n\n" +
|
|
"Within a class, a model below its minimum may take what another grew into\n" +
|
|
"under load, but not its minimum. \"default\" goes back to what the model's\n" +
|
|
"scaling implies: opportunistic when on demand, standard otherwise.",
|
|
Args: cobra.ExactArgs(2),
|
|
RunE: func(cmd *cobra.Command, args []string) error {
|
|
class := args[1]
|
|
if class == "default" {
|
|
class = ""
|
|
}
|
|
body, _ := json.Marshal(map[string]string{"class": class})
|
|
var res map[string]string
|
|
if err := client().post(cmd.Context(), "/api/v1/models/"+args[0]+"/class", body, &res); err != nil {
|
|
return err
|
|
}
|
|
fmt.Fprintf(cmd.OutOrStdout(), "%s is %s\n", res["model"], res["class"])
|
|
return nil
|
|
},
|
|
}
|
|
}
|