messages.go

// Copyright SAP SE
// SPDX-License-Identifier: Apache-2.0

package api

import (
	"errors"
	"fmt"
	"log/slog"
	"strings"

	"github.com/cobaltcore-dev/cortex/api/v1alpha1"
	"github.com/cobaltcore-dev/cortex/internal/scheduling/lib"
)

// Host object from the Nova scheduler pipeline.
// See: https://github.com/sapcc/nova/blob/stable/xena-m3/nova/scheduler/host_manager.py class HostState
type ExternalSchedulerHost struct {
	// Name of the Nova compute host, e.g. nova-compute-bb123.
	ComputeHost string `json:"host"`
	// Name of the hypervisor hostname, e.g. domain-c123.<uuid>
	HypervisorHostname string `json:"hypervisor_hostname"`
}

// Request generated by the Nova scheduler when calling cortex.
// The request contains a spec of the VM to be scheduled, a list of hosts and
// their status, and a map of weights that were calculated by the Nova weigher
// pipeline. Some additional flags are also included.
type ExternalSchedulerRequest struct {
	Spec NovaObject[NovaSpec] `json:"spec"`

	// Request context from Nova that contains additional meta information.
	Context NovaRequestContext `json:"context"`

	Hosts   []ExternalSchedulerHost `json:"hosts"`
	Weights map[string]float64      `json:"weights"`

	// The name of the pipeline to execute.
	Pipeline string `json:"pipeline"`
}

func (r ExternalSchedulerRequest) GetHosts() []string {
	hosts := make([]string, len(r.Hosts))
	for i, host := range r.Hosts {
		hosts[i] = host.ComputeHost
	}
	return hosts
}
func (r ExternalSchedulerRequest) GetWeights() map[string]float64 {
	return r.Weights
}
func (r ExternalSchedulerRequest) GetTraceLogArgs() []slog.Attr {
	greq := ""
	if r.Context.GlobalRequestID != nil {
		greq = *r.Context.GlobalRequestID
	}
	return []slog.Attr{
		slog.String("greq", greq),
		slog.String("req", r.Context.RequestID),
		slog.String("user", r.Context.UserID),
		slog.String("project", r.Context.ProjectID),
	}
}
func (r ExternalSchedulerRequest) FilterHosts(includedHosts map[string]float64) lib.FilterWeigherPipelineRequest {
	filteredHosts := make([]ExternalSchedulerHost, 0, len(includedHosts))
	for _, host := range r.Hosts {
		if _, exists := includedHosts[host.ComputeHost]; exists {
			filteredHosts = append(filteredHosts, host)
		}
	}
	r.Hosts = filteredHosts
	return r
}

type FlavorType string

const (
	// FlavorTypeGeneralPurpose represents general purpose workloads.
	FlavorTypeGeneralPurpose FlavorType = "general-purpose"
	// FlavorTypeHANA represents HANA workloads.
	FlavorTypeHANA FlavorType = "hana"
)

// GetFlavorType determines the flavor type based on the requested flavor's extra specs.
func (r ExternalSchedulerRequest) GetFlavorType() (FlavorType, error) {
	extraSpecs := r.Spec.Data.Flavor.Data.ExtraSpecs
	val, ok := extraSpecs["trait:CUSTOM_HANA_EXCLUSIVE_HOST"]
	if !ok {
		return FlavorTypeGeneralPurpose, nil
	}
	// If the key is provided, it must be either "required" or "forbidden".
	switch strings.ToLower(val) {
	case "forbidden":
		return FlavorTypeGeneralPurpose, nil
	case "required":
		return FlavorTypeHANA, nil
	default:
		return "", fmt.Errorf("unsupported value for trait:CUSTOM_HANA_EXCLUSIVE_HOST in flavor extra specs: %s", val)
	}
}

type HypervisorType string

const (
	// HypervisorTypeQEMU represents QEMU/KVM hypervisors.
	HypervisorTypeQEMU HypervisorType = "qemu"
	// HypervisorTypeCH represents Cloud-Hypervisor/KVM hypervisors.
	HypervisorTypeCH HypervisorType = "ch"
	// HypervisorTypeVMware represents VMware hypervisors.
	HypervisorTypeVMware HypervisorType = "vmware"
)

// GetHypervisorType determines the hypervisor type based on the requested flavor.
func (req ExternalSchedulerRequest) GetHypervisorType() (HypervisorType, error) {
	extraSpecs := req.Spec.Data.Flavor.Data.ExtraSpecs
	if val, ok := extraSpecs["capabilities:hypervisor_type"]; ok {
		switch strings.ToLower(val) {
		case "qemu":
			return HypervisorTypeQEMU, nil
		case "ch":
			return HypervisorTypeCH, nil
		case "vmware vcenter server":
			return HypervisorTypeVMware, nil
		}
		return "", fmt.Errorf("unsupported hypervisor_type: %s", val)
	}
	return "", errors.New("hypervisor type not specified in flavor extra specs")
}

const (
	// LiveMigrationIntent indicates that the request is intended for live migration.
	LiveMigrationIntent v1alpha1.SchedulingIntent = "live_migration"
	// RebuildIntent indicates that the request is intended for rebuilding a VM.
	RebuildIntent v1alpha1.SchedulingIntent = "rebuild"
	// ResizeIntent indicates that the request is intended for resizing a VM.
	ResizeIntent v1alpha1.SchedulingIntent = "resize"
	// EvacuateIntent indicates that the request is intended for evacuating a VM.
	EvacuateIntent v1alpha1.SchedulingIntent = "evacuate"
	// CreateIntent indicates that the request is intended for creating a new VM.
	CreateIntent v1alpha1.SchedulingIntent = "create"
	// ReserveForFailoverIntent indicates that the request is for failover reservation scheduling.
	ReserveForFailoverIntent v1alpha1.SchedulingIntent = "reserve_for_failover"
	// ReserveForCommittedResourceIntent indicates that the request is for CR reservation scheduling.
	ReserveForCommittedResourceIntent v1alpha1.SchedulingIntent = "reserve_for_committed_resource"
)

// GetIntent analyzes the request spec and determines the intent of the scheduling request.
func (req ExternalSchedulerRequest) GetIntent() (v1alpha1.SchedulingIntent, error) {
	str, err := req.Spec.Data.GetSchedulerHintStr("_nova_check_type")
	if err != nil {
		return "", err
	}
	switch str {
	// See: https://github.com/sapcc/nova/blob/ba4813/nova/scheduler/utils.py#L1251
	case "rebuild":
		return RebuildIntent, nil
	// See: https://github.com/sapcc/nova/blob/ba4813/nova/scheduler/utils.py#L1264
	case "resize":
		return ResizeIntent, nil
	// See: https://github.com/sapcc/nova/blob/ba4813/nova/scheduler/utils.py#L1277C47-L1277C63
	case "live_migrate":
		return LiveMigrationIntent, nil
	// Added in https://github.com/sapcc/nova/pull/594
	// See: https://github.com/sapcc/nova/blob/c88393/nova/compute/api.py#L5770
	case "evacuate":
		return EvacuateIntent, nil
	// Used by cortex failover reservation controller
	case "reserve_for_failover":
		return ReserveForFailoverIntent, nil
	// Used by cortex committed resource reservation controller
	case "reserve_for_committed_resource":
		return ReserveForCommittedResourceIntent, nil
	default:
		return CreateIntent, nil
	}
}

// Response generated by cortex for the Nova scheduler.
// Cortex returns an ordered list of hosts that the VM should be scheduled on.
type ExternalSchedulerResponse struct {
	Hosts []string `json:"hosts"`
}

// Wrapped Nova object. Nova returns objects in this format.
type NovaObject[V any] struct {
	Name      string   `json:"nova_object.name"`
	Namespace string   `json:"nova_object.namespace"`
	Version   string   `json:"nova_object.version"`
	Data      V        `json:"nova_object.data"`
	Changes   []string `json:"nova_object.changes"`
}

type NovaObjectList[V any] struct {
	Objects []NovaObject[V] `json:"objects"`
}

// Spec object from the Nova scheduler pipeline. Fields unused by Nova are omitted.
// See: https://github.com/sapcc/nova/blob/stable/xena-m3/nova/objects/request_spec.py
type NovaSpec struct {
	ProjectID        string `json:"project_id"`
	UserID           string `json:"user_id"`
	InstanceUUID     string `json:"instance_uuid"`
	AvailabilityZone string `json:"availability_zone"`
	NumInstances     uint64 `json:"num_instances"`
	IsBfv            bool   `json:"is_bfv"`

	// Consider using GetSchedulerHintStr.
	SchedulerHints map[string]any `json:"scheduler_hints"`

	IgnoreHosts *[]string `json:"ignore_hosts"`
	ForceHosts  *[]string `json:"force_hosts"`
	ForceNodes  *[]string `json:"force_nodes"`

	Image              NovaObject[NovaImageMeta]          `json:"image"`
	Flavor             NovaObject[NovaFlavor]             `json:"flavor"`
	RequestLevelParams NovaObject[NovaRequestLevelParams] `json:"request_level_params"`
	NetworkMetadata    NovaObject[map[string]any]         `json:"network_metadata"`
	Limits             NovaObject[map[string]any]         `json:"limits"`
	RequestedNetworks  NovaObjectList[map[string]any]     `json:"requested_networks"`
	SecurityGroups     NovaObjectList[map[string]any]     `json:"security_groups"`

	NumaTopology         *NovaObject[NovaNumaTopology]         `json:"numa_topology"`
	RequestedDestination *NovaObject[NovaRequestedDestination] `json:"requested_destination"`
	InstanceGroup        *NovaObject[NovaInstanceGroup]        `json:"instance_group"`
}

// Hints can be a one-element list or a direct value.
// See: https://github.com/sapcc/nova/blob/3e715db/nova/objects/request_spec.py#L382
//
// This function is a convenience to extract the first element
// from the hints list or return the value directly if it's not a list.
func (s NovaSpec) GetSchedulerHintStr(key string) (string, error) {
	if s.SchedulerHints == nil {
		return "", errors.New("scheduler hints are not set")
	}
	raw, ok := s.SchedulerHints[key]
	if !ok {
		return "", fmt.Errorf("scheduler hint %q not found", key)
	}
	switch v := raw.(type) {
	case []any:
		if len(v) > 0 {
			if str, ok := v[0].(string); ok {
				return str, nil
			}
		}
	case string:
		return v, nil
	}
	return "", errors.New("unknown scheduler hint type")
}

type NovaInstanceGroup struct {
	UserID    string         `json:"user_id"`
	ProjectID string         `json:"project_id"`
	UUID      string         `json:"uuid"`
	Name      string         `json:"name"`
	Policies  []string       `json:"policies"`
	Members   []string       `json:"members"`
	Hosts     []string       `json:"hosts"`
	Policy    string         `json:"policy"`
	Rules     map[string]any `json:"_rules"`
	CreatedAt string         `json:"created_at"`
	UpdatedAt *string        `json:"updated_at"`
	DeletedAt *string        `json:"deleted_at"`
	Deleted   bool           `json:"deleted"`
}

// Nova image metadata for the specified VM.
type NovaImageMeta struct {
	ID              string                     `json:"id"`
	Name            string                     `json:"name"`
	Status          string                     `json:"status"`
	Checksum        string                     `json:"checksum"`
	Owner           string                     `json:"owner"`
	Size            int                        `json:"size"`
	ContainerFormat string                     `json:"container_format"`
	DiskFormat      string                     `json:"disk_format"`
	CreatedAt       string                     `json:"created_at"`
	UpdatedAt       string                     `json:"updated_at"`
	MinRAM          int                        `json:"min_ram"`
	MinDisk         int                        `json:"min_disk"`
	Properties      NovaObject[map[string]any] `json:"properties"`
}

// Nova flavor metadata for the specified VM.
type NovaFlavor struct {
	ID          int               `json:"id"`
	Name        string            `json:"name"`
	MemoryMB    uint64            `json:"memory_mb"`
	VCPUs       uint64            `json:"vcpus"`
	RootGB      uint64            `json:"root_gb"`
	EphemeralGB uint64            `json:"ephemeral_gb"`
	FlavorID    string            `json:"flavorid"`
	Swap        int               `json:"swap"`
	RXTXFactor  float64           `json:"rxtx_factor"`
	VCPUWeight  int               `json:"vcpu_weight"`
	Disabled    bool              `json:"disabled"`
	IsPublic    bool              `json:"is_public"`
	ExtraSpecs  map[string]string `json:"extra_specs"`
	Description *string           `json:"description"`
	CreatedAt   string            `json:"created_at"`
	UpdatedAt   *string           `json:"updated_at"`
	DeletedAt   *string           `json:"deleted_at"`
	Deleted     bool              `json:"deleted"`
}

// Nova NUMA topology for the specified VM.
type NovaNumaTopology struct {
	Cells []NovaObject[map[string]any] `json:"cells"`
}

type NovaRequestedDestination struct {
	Host                string    `json:"host"`
	Node                string    `json:"node"`
	Aggregates          []string  `json:"aggregates"`
	ForbiddenAggregates *[]string `json:"forbidden_aggregates"`
}

type NovaRequestLevelParams struct {
	RootRequired  []any `json:"root_required"`
	RootForbidden []any `json:"root_forbidden"`
	SameSubtree   []any `json:"same_subtree"`
}

// Nova request context object. For the spec of this object, see:
//
// - This: https://github.com/sapcc/nova/blob/a56409/nova/context.py#L166
// - And: https://github.com/openstack/oslo.context/blob/db20dd/oslo_context/context.py#L329
//
// Some fields are omitted: "service_catalog", "read_deleted" (same as "show_deleted")
type NovaRequestContext struct {
	// Fields added by oslo.context

	UserID          string   `json:"user"`
	ProjectID       string   `json:"project_id"`
	SystemScope     *string  `json:"system_scope"`
	Project         string   `json:"project"`
	DomainID        *string  `json:"domain"`
	UserDomainID    string   `json:"user_domain"`
	ProjectDomainID string   `json:"project_domain"`
	IsAdmin         bool     `json:"is_admin"`
	ReadOnly        bool     `json:"read_only"`
	ShowDeleted     bool     `json:"show_deleted"`
	RequestID       string   `json:"request_id"`
	GlobalRequestID *string  `json:"global_request_id"`
	ResourceUUID    *string  `json:"resource_uuid"`
	Roles           []string `json:"roles"`
	UserIdentity    string   `json:"user_identity"`
	IsAdminProject  bool     `json:"is_admin_project"`
	ReadDeleted     string   `json:"read_deleted"`

	// Fields added by the Nova scheduler

	RemoteAddress string  `json:"remote_address"`
	Timestamp     string  `json:"timestamp"`
	QuotaClass    *string `json:"quota_class"`
	UserName      string  `json:"user_name"`
	ProjectName   string  `json:"project_name"`
}