Datacenter no more handle booking but is fully charged with Kube & minio allocate per NATS

This commit is contained in:
mr
2026-02-25 09:08:40 +01:00
parent 41750dc054
commit b45c795002
19 changed files with 1494 additions and 2109 deletions

View File

@@ -0,0 +1,219 @@
package minio
import (
"context"
"encoding/json"
"fmt"
"oc-datacenter/conf"
oclib "cloud.o-forge.io/core/oc-lib"
"github.com/minio/madmin-go/v4"
"github.com/minio/minio-go/v7"
"github.com/minio/minio-go/v7/pkg/credentials"
v1 "k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/client-go/kubernetes"
"k8s.io/client-go/rest"
"github.com/necmettindev/randomstring"
)
type MinioService struct {
Url string
RootKey string
RootSecret string
MinioAdminClient *madmin.AdminClient
}
type StatementEntry struct {
Effect string `json:"Effect"`
Action []string `json:"Action"`
Resource string `json:"Resource"`
}
type PolicyDocument struct {
Version string `json:"Version"`
Statement []StatementEntry `json:"Statement"`
}
func NewMinioService(url string) *MinioService {
return &MinioService{
Url: url,
RootKey: conf.GetConfig().MinioRootKey,
RootSecret: conf.GetConfig().MinioRootSecret,
}
}
func (m *MinioService) CreateClient() error {
cred := credentials.NewStaticV4(m.RootKey, m.RootSecret, "")
cli, err := madmin.NewWithOptions(m.Url, &madmin.Options{Creds: cred, Secure: false}) // Maybe in the future we should use the secure option ?
if err != nil {
return err
}
m.MinioAdminClient = cli
return nil
}
func (m *MinioService) CreateCredentials(executionId string) (string, string, error) {
policy := PolicyDocument{
Version: "2012-10-17",
Statement: []StatementEntry{
{
Effect: "Allow",
Action: []string{"s3:GetObject", "s3:PutObject"},
Resource: "arn:aws:s3:::" + executionId + "/*",
},
},
}
p, err := json.Marshal(policy)
if err != nil {
return "", "", err
}
randAccess, randSecret := getRandomCreds()
req := madmin.AddServiceAccountReq{
Policy: p,
TargetUser: m.RootKey,
AccessKey: randAccess,
SecretKey: randSecret,
}
res, err := m.MinioAdminClient.AddServiceAccount(context.Background(), req)
if err != nil {
return "", "", err
}
return res.AccessKey, res.SecretKey, nil
}
func getRandomCreds() (string, string) {
opts := randomstring.GenerationOptions{
Length: 20,
}
a, _ := randomstring.GenerateString(opts)
opts.Length = 40
s, _ := randomstring.GenerateString(opts)
return a, s
}
func (m *MinioService) CreateMinioConfigMap(minioID string, executionId string, url string) error {
config, err := rest.InClusterConfig()
if err != nil {
return err
}
clientset, err := kubernetes.NewForConfig(config)
if err != nil {
return err
}
configMap := &v1.ConfigMap{
ObjectMeta: metav1.ObjectMeta{
Name: minioID + "artifact-repository",
Namespace: executionId,
},
Data: map[string]string{
minioID + "s3-local": fmt.Sprintf(`
s3:
bucket: %s
endpoint: %s
insecure: true
accessKeySecret:
name: %s-secret-s3
key: accesskey
secretKeySecret:
name: %s-secret-s3
key: secretkey
`, minioID+"-"+executionId, url, minioID, minioID),
},
}
existing, err := clientset.CoreV1().
ConfigMaps(executionId).
Get(context.Background(), minioID+"artifact-repository", metav1.GetOptions{})
if err == nil {
// Update
existing.Data = configMap.Data
_, err = clientset.CoreV1().
ConfigMaps(executionId).
Update(context.Background(), existing, metav1.UpdateOptions{})
} else {
// Create
_, err = clientset.CoreV1().
ConfigMaps(executionId).
Create(context.Background(), configMap, metav1.CreateOptions{})
}
return nil
}
func (m *MinioService) CreateBucket(minioID string, executionId string) error {
l := oclib.GetLogger()
cred := credentials.NewStaticV4(m.RootKey, m.RootSecret, "")
client, err := minio.New(m.Url, &minio.Options{
Creds: cred,
Secure: false,
})
if err != nil {
l.Error().Msg("Error when creating the minio client for the data plane")
return err
}
err = client.MakeBucket(context.Background(), minioID+"-"+executionId, minio.MakeBucketOptions{})
if err != nil {
l.Error().Msg("Error when creating the bucket for namespace " + executionId)
return err
}
l.Info().Msg("Created the bucket " + minioID + "-" + executionId + " on " + m.Url + " minio")
return nil
}
// DeleteCredentials revokes a scoped Minio service account by its access key.
func (m *MinioService) DeleteCredentials(accessKey string) error {
if err := m.MinioAdminClient.DeleteServiceAccount(context.Background(), accessKey); err != nil {
return fmt.Errorf("DeleteCredentials: %w", err)
}
return nil
}
// DeleteBucket removes the execution bucket from Minio.
func (m *MinioService) DeleteBucket(minioID, executionId string) error {
l := oclib.GetLogger()
cred := credentials.NewStaticV4(m.RootKey, m.RootSecret, "")
client, err := minio.New(m.Url, &minio.Options{Creds: cred, Secure: false})
if err != nil {
l.Error().Msg("Error when creating minio client for bucket deletion")
return err
}
bucketName := minioID + "-" + executionId
if err := client.RemoveBucket(context.Background(), bucketName); err != nil {
l.Error().Msg("Error when deleting bucket " + bucketName)
return err
}
l.Info().Msg("Deleted bucket " + bucketName + " on " + m.Url)
return nil
}
// DeleteMinioConfigMap removes the artifact-repository ConfigMap from the execution namespace.
func (m *MinioService) DeleteMinioConfigMap(minioID, executionId string) error {
cfg, err := rest.InClusterConfig()
if err != nil {
return err
}
clientset, err := kubernetes.NewForConfig(cfg)
if err != nil {
return err
}
return clientset.CoreV1().ConfigMaps(executionId).Delete(
context.Background(), minioID+"artifact-repository", metav1.DeleteOptions{},
)
}

View File

@@ -0,0 +1,297 @@
package minio
import (
"context"
"encoding/json"
"fmt"
"slices"
"oc-datacenter/conf"
oclib "cloud.o-forge.io/core/oc-lib"
"cloud.o-forge.io/core/oc-lib/models/live"
"cloud.o-forge.io/core/oc-lib/tools"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
)
// MinioCredentialEvent is the NATS payload used to transfer Minio credentials between peers.
//
// Two-phase protocol over PROPALGATION_EVENT (Action = PB_MINIO_CONFIG):
// - Phase 1 role assignment (Access == ""):
// oc-discovery routes this to the SOURCE peer (Minio host) → InitializeAsSource.
// - Phase 2 credential delivery (Access != ""):
// oc-discovery routes this to the TARGET peer (compute host) → InitializeAsTarget.
type MinioCredentialEvent struct {
ExecutionsID string `json:"executions_id"`
MinioID string `json:"minio_id"`
Access string `json:"access"`
Secret string `json:"secret"`
SourcePeerID string `json:"source_peer_id"`
DestPeerID string `json:"dest_peer_id"`
URL string `json:"url"`
// OriginID is the peer that initiated the provisioning request.
// The PB_CONSIDERS response is routed back to this peer.
OriginID string `json:"origin_id"`
}
// minioConsidersPayload is the PB_CONSIDERS payload emitted after minio provisioning.
type minioConsidersPayload struct {
OriginID string `json:"origin_id"`
ExecutionsID string `json:"executions_id"`
Secret string `json:"secret,omitempty"`
Error *string `json:"error,omitempty"`
}
// emitConsiders publishes a PB_CONSIDERS back to OriginID with the result of
// the minio provisioning. secret is the provisioned credential; err is nil on success.
func emitConsiders(executionsID, originID, secret string, provErr error) {
var errStr *string
if provErr != nil {
s := provErr.Error()
errStr = &s
}
payload, _ := json.Marshal(minioConsidersPayload{
OriginID: originID,
ExecutionsID: executionsID,
Secret: secret,
Error: errStr,
})
b, _ := json.Marshal(&tools.PropalgationMessage{
DataType: tools.STORAGE_RESOURCE.EnumIndex(),
Action: tools.PB_CONSIDERS,
Payload: payload,
})
go tools.NewNATSCaller().SetNATSPub(tools.PROPALGATION_EVENT, tools.NATSResponse{
FromApp: "oc-datacenter",
Datatype: -1,
Method: int(tools.PROPALGATION_EVENT),
Payload: b,
})
}
// MinioSetter carries the execution context for a Minio credential provisioning.
type MinioSetter struct {
ExecutionsID string // used as both the bucket name and the K8s namespace suffix
MinioID string // ID of the Minio storage resource
}
func NewMinioSetter(execID, minioID string) *MinioSetter {
return &MinioSetter{ExecutionsID: execID, MinioID: minioID}
}
// InitializeAsSource is called on the peer that hosts the Minio instance.
//
// It:
// 1. Looks up the live-storage endpoint URL for MinioID.
// 2. Creates a scoped service account (access + secret limited to the execution bucket).
// 3. Creates the execution bucket.
// 4. If source and dest are the same peer, calls InitializeAsTarget directly.
// Otherwise, publishes a MinioCredentialEvent via NATS (Phase 2) so that
// oc-discovery can route the credentials to the compute peer.
func (m *MinioSetter) InitializeAsSource(ctx context.Context, localPeerID, destPeerID, originID string) {
logger := oclib.GetLogger()
url, err := m.loadMinioURL(localPeerID)
if err != nil {
logger.Error().Msg("MinioSetter.InitializeAsSource: " + err.Error())
return
}
service := NewMinioService(url)
if err := service.CreateClient(); err != nil {
logger.Error().Msg("MinioSetter.InitializeAsSource: failed to create admin client: " + err.Error())
return
}
access, secret, err := service.CreateCredentials(m.ExecutionsID)
if err != nil {
logger.Error().Msg("MinioSetter.InitializeAsSource: failed to create service account: " + err.Error())
return
}
if err := service.CreateBucket(m.MinioID, m.ExecutionsID); err != nil {
logger.Error().Msg("MinioSetter.InitializeAsSource: failed to create bucket: " + err.Error())
return
}
logger.Info().Msg("MinioSetter.InitializeAsSource: bucket and service account ready for " + m.ExecutionsID)
event := MinioCredentialEvent{
ExecutionsID: m.ExecutionsID,
MinioID: m.MinioID,
Access: access,
Secret: secret,
SourcePeerID: localPeerID,
DestPeerID: destPeerID,
OriginID: originID,
}
if destPeerID == localPeerID {
// Same peer: store the secret locally without going through NATS.
m.InitializeAsTarget(ctx, event)
return
}
// Cross-peer: publish credentials (Phase 2) so oc-discovery routes them to the compute peer.
payload, err := json.Marshal(event)
if err != nil {
logger.Error().Msg("MinioSetter.InitializeAsSource: failed to marshal credential event: " + err.Error())
return
}
if b, err := json.Marshal(&tools.PropalgationMessage{
DataType: -1,
Action: tools.PB_MINIO_CONFIG,
Payload: payload,
}); err == nil {
go tools.NewNATSCaller().SetNATSPub(tools.PROPALGATION_EVENT, tools.NATSResponse{
FromApp: "oc-datacenter",
Datatype: -1,
User: "",
Method: int(tools.PROPALGATION_EVENT),
Payload: b,
})
logger.Info().Msg("MinioSetter.InitializeAsSource: credentials published via NATS for " + m.ExecutionsID)
}
}
// InitializeAsTarget is called on the peer that runs the compute workload.
//
// It stores the Minio credentials received from the source peer (via NATS or directly)
// as a Kubernetes secret inside the execution namespace, making them available to pods.
func (m *MinioSetter) InitializeAsTarget(ctx context.Context, event MinioCredentialEvent) {
logger := oclib.GetLogger()
k, err := tools.NewKubernetesService(
conf.GetConfig().KubeHost+":"+conf.GetConfig().KubePort,
conf.GetConfig().KubeCA, conf.GetConfig().KubeCert, conf.GetConfig().KubeData,
)
if err != nil {
logger.Error().Msg("MinioSetter.InitializeAsTarget: failed to create k8s service: " + err.Error())
return
}
if err := k.CreateSecret(ctx, event.MinioID, event.ExecutionsID, event.Access, event.Secret); err != nil {
logger.Error().Msg("MinioSetter.InitializeAsTarget: failed to create k8s secret: " + err.Error())
emitConsiders(event.ExecutionsID, event.OriginID, "", err)
return
}
if err := NewMinioService(event.URL).CreateMinioConfigMap(event.MinioID, event.ExecutionsID, event.URL); err == nil {
logger.Error().Msg("MinioSetter.InitializeAsTarget: failed to create config map: " + err.Error())
emitConsiders(event.ExecutionsID, event.OriginID, "", err)
return
}
logger.Info().Msg("MinioSetter.InitializeAsTarget: Minio credentials stored in namespace " + event.ExecutionsID)
emitConsiders(event.ExecutionsID, event.OriginID, event.Secret, nil)
}
// MinioDeleteEvent is the NATS payload used to tear down Minio resources.
// It mirrors MinioCredentialEvent but carries the access key for revocation.
type MinioDeleteEvent struct {
ExecutionsID string `json:"executions_id"`
MinioID string `json:"minio_id"`
Access string `json:"access"` // service account access key to revoke on the Minio host
SourcePeerID string `json:"source_peer_id"`
DestPeerID string `json:"dest_peer_id"`
OriginID string `json:"origin_id"`
}
// TeardownAsTarget is called on the peer that runs the compute workload.
// It reads the stored access key from the K8s secret, then removes both the secret
// and the artifact-repository ConfigMap from the execution namespace.
// For same-peer deployments it calls TeardownAsSource directly; otherwise it
// publishes a MinioDeleteEvent via NATS (PB_DELETE) so oc-discovery routes it to
// the Minio host peer.
func (m *MinioSetter) TeardownAsTarget(ctx context.Context, event MinioDeleteEvent) {
logger := oclib.GetLogger()
k, err := tools.NewKubernetesService(
conf.GetConfig().KubeHost+":"+conf.GetConfig().KubePort,
conf.GetConfig().KubeCA, conf.GetConfig().KubeCert, conf.GetConfig().KubeData,
)
if err != nil {
logger.Error().Msg("MinioSetter.TeardownAsTarget: failed to create k8s service: " + err.Error())
emitConsiders(event.ExecutionsID, event.OriginID, "", err)
return
}
// Read the access key from the K8s secret before deleting it.
accessKey := event.Access
if accessKey == "" {
if secret, err := k.Set.CoreV1().Secrets(event.ExecutionsID).Get(
ctx, event.MinioID+"-secret-s3", metav1.GetOptions{},
); err == nil {
accessKey = string(secret.Data["access-key"])
}
}
// Delete K8s credentials secret.
if err := k.Set.CoreV1().Secrets(event.ExecutionsID).Delete(
ctx, event.MinioID+"-secret-s3", metav1.DeleteOptions{},
); err != nil {
logger.Error().Msg("MinioSetter.TeardownAsTarget: failed to delete secret: " + err.Error())
}
// Delete artifact-repository ConfigMap.
if err := NewMinioService("").DeleteMinioConfigMap(event.MinioID, event.ExecutionsID); err != nil {
logger.Error().Msg("MinioSetter.TeardownAsTarget: failed to delete configmap: " + err.Error())
}
logger.Info().Msg("MinioSetter.TeardownAsTarget: K8s resources removed for " + event.ExecutionsID)
// For same-peer deployments the source cleanup runs directly here so the
// caller (REMOVE_EXECUTION handler) doesn't have to distinguish roles.
if event.SourcePeerID == event.DestPeerID {
event.Access = accessKey
m.TeardownAsSource(ctx, event)
}
}
// TeardownAsSource is called on the peer that hosts the Minio instance.
// It revokes the scoped service account and removes the execution bucket.
func (m *MinioSetter) TeardownAsSource(ctx context.Context, event MinioDeleteEvent) {
logger := oclib.GetLogger()
url, err := m.loadMinioURL(event.SourcePeerID)
if err != nil {
logger.Error().Msg("MinioSetter.TeardownAsSource: " + err.Error())
return
}
svc := NewMinioService(url)
if err := svc.CreateClient(); err != nil {
logger.Error().Msg("MinioSetter.TeardownAsSource: failed to create admin client: " + err.Error())
return
}
if event.Access != "" {
if err := svc.DeleteCredentials(event.Access); err != nil {
logger.Error().Msg("MinioSetter.TeardownAsSource: failed to delete service account: " + err.Error())
}
}
if err := svc.DeleteBucket(event.MinioID, event.ExecutionsID); err != nil {
logger.Error().Msg("MinioSetter.TeardownAsSource: failed to delete bucket: " + err.Error())
}
logger.Info().Msg("MinioSetter.TeardownAsSource: Minio resources removed for " + event.ExecutionsID)
}
// loadMinioURL searches through all live storages accessible by peerID to find
// the one that references MinioID, and returns its endpoint URL.
func (m *MinioSetter) loadMinioURL(peerID string) (string, error) {
res := oclib.NewRequest(oclib.LibDataEnum(oclib.LIVE_STORAGE), "", peerID, []string{}, nil).LoadAll(false)
if res.Err != "" {
return "", fmt.Errorf("loadMinioURL: failed to load live storages: %s", res.Err)
}
for _, dbo := range res.Data {
l := dbo.(*live.LiveStorage)
if slices.Contains(l.ResourcesID, m.MinioID) {
return l.Source, nil
}
}
return "", fmt.Errorf("loadMinioURL: no live storage found for minio ID %s", m.MinioID)
}