Jon Chery
41bcf0a6bf
orca node drain <host>: marks draining, stops allocs via SSH, marks drained. orca daemon drain-and-stop: stops v0.8 daemons on peers. orca job migrate <name> --to <node>: drain+reschedule composite (C3=a, not live-migrate). Node states: draining, drained. ---ci--- project: orca phase: 05 milestone: v0.11 status: execute ---/ci---
84 lines
2.6 KiB
Go
84 lines
2.6 KiB
Go
package engine
|
|
|
|
import (
|
|
"context"
|
|
"fmt"
|
|
"log/slog"
|
|
|
|
"git.cloudinit.dev/coreci/orca/internal/model"
|
|
"git.cloudinit.dev/coreci/orca/internal/store"
|
|
)
|
|
|
|
type NodeRegistry struct {
|
|
repo *store.NodeRepo
|
|
audit *Audit
|
|
log *slog.Logger
|
|
}
|
|
|
|
func NewNodeRegistry(repo *store.NodeRepo, audit *Audit, log *slog.Logger) *NodeRegistry {
|
|
if log == nil {
|
|
log = slog.Default()
|
|
}
|
|
return &NodeRegistry{repo: repo, audit: audit, log: log}
|
|
}
|
|
|
|
func (r *NodeRegistry) Join(ctx context.Context, n *model.Node) error {
|
|
if err := r.repo.Insert(ctx, n); err != nil {
|
|
r.audit.Record(ctx, "cli", "node.join", n.ID, "failure", err, map[string]any{
|
|
"name": n.Name,
|
|
"address": n.Address,
|
|
})
|
|
return fmt.Errorf("join node: %w", err)
|
|
}
|
|
r.audit.Record(ctx, "cli", "node.join", n.ID, "success", nil, map[string]any{
|
|
"name": n.Name,
|
|
"address": n.Address,
|
|
})
|
|
r.log.Info("node joined",
|
|
slog.String("node_id", n.ID),
|
|
slog.String("name", n.Name),
|
|
slog.String("address", n.Address))
|
|
return nil
|
|
}
|
|
|
|
func (r *NodeRegistry) Leave(ctx context.Context, id string) error {
|
|
if err := r.repo.UpdateState(ctx, id, model.NodeStateLeft); err != nil {
|
|
r.audit.Record(ctx, "cli", "node.leave", id, "failure", err, nil)
|
|
return fmt.Errorf("leave node: %w", err)
|
|
}
|
|
r.audit.Record(ctx, "cli", "node.leave", id, "success", nil, nil)
|
|
r.log.Info("node left", slog.String("node_id", id))
|
|
return nil
|
|
}
|
|
|
|
func (r *NodeRegistry) Forget(ctx context.Context, id string) error {
|
|
if err := r.repo.Delete(ctx, id); err != nil {
|
|
r.audit.Record(ctx, "cli", "node.forget", id, "failure", err, nil)
|
|
return fmt.Errorf("forget node: %w", err)
|
|
}
|
|
r.audit.Record(ctx, "cli", "node.forget", id, "success", nil, nil)
|
|
r.log.Info("node removed from registry", slog.String("node_id", id))
|
|
return nil
|
|
}
|
|
|
|
func (r *NodeRegistry) List(ctx context.Context) ([]*model.Node, error) {
|
|
return r.repo.List(ctx)
|
|
}
|
|
|
|
func (r *NodeRegistry) Get(ctx context.Context, id string) (*model.Node, error) {
|
|
return r.repo.Get(ctx, id)
|
|
}
|
|
|
|
// SetNodeState updates a node's state to the given raw string (REQ-061).
|
|
// Used by the drain commands for the draining/drained states. This is
|
|
// the registry-level wrapper around store.NodeRepo.SetNodeState so the
|
|
// cli package does not need to reach into the repo directly.
|
|
func (r *NodeRegistry) SetNodeState(ctx context.Context, id, state string) error {
|
|
if err := r.repo.SetNodeState(ctx, id, state); err != nil {
|
|
r.audit.Record(ctx, "cli", "node.set_state", id, "failure", err, map[string]any{"state": state})
|
|
return fmt.Errorf("set node state: %w", err)
|
|
}
|
|
r.log.Info("node state set", slog.String("node_id", id), slog.String("state", state))
|
|
return nil
|
|
}
|