Move rules deletion from Purger to Ruler API. (#3899)

Signed-off-by: Peter Štibraný <[email protected]>

Author: pstibrany

CHANGELOG.md

@@ -72,7 +72,7 @@
   * `cortex_ha_tracker_replicas_cleanup_marked_for_deletion_total`
   * `cortex_ha_tracker_replicas_cleanup_deleted_total`
   * `cortex_ha_tracker_replicas_cleanup_delete_failed_total`
-* [ENHANCEMENT] Tenant deletion endpoints now support deletion of ruler groups. This only works when using rule store that supports deletion. #3750
+* [ENHANCEMENT] Ruler now has new API endpoint `/ruler/delete_tenant_config` that can be used to delete all ruler groups for tenant. It is intended to be used by administrators who wish to clean up state after removed user. Note that this endpoint is enabled regardless of `-experimental.ruler.enable-api`. #3750 #3899
 * [ENHANCEMENT] Query-frontend, query-scheduler: cleanup metrics for inactive tenants. #3826
 * [ENHANCEMENT] Distributor: Prevent failed ingestion from affecting rate limiting. #3825
 * [ENHANCEMENT] Blocks storage: added `-blocks-storage.s3.region` support to S3 client configuration. #3811

docs/api/_index.md

@@ -49,6 +49,7 @@ For the sake of clarity, in this document we have grouped API endpoints by servi
 | [Set rule group](#set-rule-group) | Ruler | `POST /api/v1/rules/{namespace}` |
 | [Delete rule group](#delete-rule-group) | Ruler | `DELETE /api/v1/rules/{namespace}/{groupName}` |
 | [Delete namespace](#delete-namespace) | Ruler | `DELETE /api/v1/rules/{namespace}` |
+| [Delete tenant configuration](#delete-tenant-configuration) | Ruler | `POST /ruler/delete_tenant_config` |
 | [Alertmanager status](#alertmanager-status) | Alertmanager | `GET /multitenant_alertmanager/status` |
 | [Alertmanager ring status](#alertmanager-ring-status) | Alertmanager | `GET /multitenant_alertmanager/ring` |
 | [Alertmanager UI](#alertmanager-ui) | Alertmanager | `GET /<alertmanager-http-prefix>` |
@@ -636,6 +637,18 @@ _This experimental endpoint is disabled by default and can be enabled via the `-
 
 _Requires [authentication](#authentication)._
 
+### Delete tenant configuration
+
+```
+POST /ruler/delete_tenant_config
+```
+
+This deletes all rule groups for tenant, and returns `200` on success. Calling endpoint when no rule groups exist for user returns `200`. Authentication is only to identify the tenant.
+
+This is intended as internal API, and not to be exposed to users. This endpoint is enabled regardless of whether `-experimental.ruler.enable-api` is enabled or not.
+
+_Requires [authentication](#authentication)._
+
 ## Alertmanager
 
 ### Alertmanager status

pkg/api/api.go

@@ -276,6 +276,9 @@ func (a *API) RegisterRuler(r *ruler.Ruler) {
 	a.indexPage.AddLink(SectionAdminEndpoints, "/ruler/ring", "Ruler Ring Status")
 	a.RegisterRoute("/ruler/ring", r, false, "GET", "POST")
 
+	// Administrative API, uses authentication to inform which user's configuration to delete.
+	a.RegisterRoute("/ruler/delete_tenant_config", http.HandlerFunc(r.DeleteTenantConfiguration), true, "POST")
+
 	// Legacy Ring Route
 	a.RegisterRoute("/ruler_ring", r, false, "GET", "POST")
 

pkg/chunk/purger/tenant_deletion_api.go

@@ -13,7 +13,6 @@ import (
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/thanos-io/thanos/pkg/objstore"
 
-	"github.com/cortexproject/cortex/pkg/ruler/rulestore"
 	"github.com/cortexproject/cortex/pkg/storage/bucket"
 	cortex_tsdb "github.com/cortexproject/cortex/pkg/storage/tsdb"
 	"github.com/cortexproject/cortex/pkg/tenant"
@@ -22,24 +21,22 @@ import (
 
 type TenantDeletionAPI struct {
 	bucketClient objstore.Bucket
-	ruleStore    rulestore.RuleStore
 	logger       log.Logger
 	cfgProvider  bucket.TenantConfigProvider
 }
 
-func NewTenantDeletionAPI(storageCfg cortex_tsdb.BlocksStorageConfig, cfgProvider bucket.TenantConfigProvider, ruleStore rulestore.RuleStore, logger log.Logger, reg prometheus.Registerer) (*TenantDeletionAPI, error) {
+func NewTenantDeletionAPI(storageCfg cortex_tsdb.BlocksStorageConfig, cfgProvider bucket.TenantConfigProvider, logger log.Logger, reg prometheus.Registerer) (*TenantDeletionAPI, error) {
 	bucketClient, err := createBucketClient(storageCfg, logger, reg)
 	if err != nil {
 		return nil, err
 	}
 
-	return newTenantDeletionAPI(bucketClient, cfgProvider, ruleStore, logger), nil
+	return newTenantDeletionAPI(bucketClient, cfgProvider, logger), nil
 }
 
-func newTenantDeletionAPI(bkt objstore.Bucket, cfgProvider bucket.TenantConfigProvider, ruleStore rulestore.RuleStore, logger log.Logger) *TenantDeletionAPI {
+func newTenantDeletionAPI(bkt objstore.Bucket, cfgProvider bucket.TenantConfigProvider, logger log.Logger) *TenantDeletionAPI {
 	return &TenantDeletionAPI{
 		bucketClient: bkt,
-		ruleStore:    ruleStore,
 		cfgProvider:  cfgProvider,
 		logger:       logger,
 	}
@@ -49,7 +46,9 @@ func (api *TenantDeletionAPI) DeleteTenant(w http.ResponseWriter, r *http.Reques
 	ctx := r.Context()
 	userID, err := tenant.TenantID(ctx)
 	if err != nil {
-		http.Error(w, err.Error(), http.StatusBadRequest)
+		// When Cortex is running, it uses Auth Middleware for checking X-Scope-OrgID and injecting tenant into context.
+		// Auth Middleware sends http.StatusUnauthorized if X-Scope-OrgID is missing, so we do too here, for consistency.
+		http.Error(w, err.Error(), http.StatusUnauthorized)
 		return
 	}
 
@@ -63,38 +62,12 @@ func (api *TenantDeletionAPI) DeleteTenant(w http.ResponseWriter, r *http.Reques
 
 	level.Info(api.logger).Log("msg", "tenant deletion mark in blocks storage created", "user", userID)
 
-	if api.ruleStore != nil {
-		err := api.deleteRules(r.Context(), userID)
-		if err != nil {
-			level.Error(api.logger).Log("msg", "failed to delete tenant rule groups", "user", userID, "err", err)
-			http.Error(w, errors.Wrapf(err, "failed to delete tenant rule groups").Error(), http.StatusInternalServerError)
-			return
-		}
-	}
-
 	w.WriteHeader(http.StatusOK)
 }
 
-func (api *TenantDeletionAPI) deleteRules(ctx context.Context, userID string) error {
-	if !api.ruleStore.SupportsModifications() {
-		level.Warn(api.logger).Log("msg", "cannot delete tenant rule groups, using read-only rule store", "user", userID)
-		return nil
-	}
-
-	err := api.ruleStore.DeleteNamespace(ctx, userID, "") // Empty namespace = delete all rule groups.
-	if err != nil && !errors.Is(err, rulestore.ErrGroupNamespaceNotFound) {
-		return err
-	}
-
-	level.Info(api.logger).Log("msg", "deleted all tenant rule groups", "user", userID)
-	return nil
-}
-
 type DeleteTenantStatusResponse struct {
-	TenantID                  string `json:"tenant_id"`
-	BlocksDeleted             bool   `json:"blocks_deleted"`
-	RuleGroupsDeleted         bool   `json:"rule_groups_deleted"`
-	AlertManagerConfigDeleted bool   `json:"alert_manager_config_deleted,omitempty"` // Not yet supported.
+	TenantID      string `json:"tenant_id"`
+	BlocksDeleted bool   `json:"blocks_deleted"`
 }
 
 func (api *TenantDeletionAPI) DeleteTenantStatus(w http.ResponseWriter, r *http.Request) {
@@ -113,29 +86,9 @@ func (api *TenantDeletionAPI) DeleteTenantStatus(w http.ResponseWriter, r *http.
 		return
 	}
 
-	result.RuleGroupsDeleted, err = api.isRulesForUserDeleted(ctx, userID)
-	if err != nil {
-		http.Error(w, err.Error(), http.StatusInternalServerError)
-		return
-	}
-
 	util.WriteJSONResponse(w, result)
 }
 
-func (api *TenantDeletionAPI) isRulesForUserDeleted(ctx context.Context, userID string) (bool, error) {
-	if api.ruleStore == nil {
-		// If API doesn't have access to rule store, then we cannot say that rules have been deleted.
-		return false, nil
-	}
-
-	list, err := api.ruleStore.ListRuleGroupsForUserAndNamespace(ctx, userID, "")
-	if err != nil {
-		return false, errors.Wrap(err, "failed to list rule groups for tenant")
-	}
-
-	return len(list) == 0, nil
-}
-
 func (api *TenantDeletionAPI) isBlocksForUserDeleted(ctx context.Context, userID string) (bool, error) {
 	var errBlockFound = errors.New("block found")
 

pkg/chunk/purger/tenant_deletion_api_test.go

@@ -3,33 +3,27 @@ package purger
 import (
 	"bytes"
 	"context"
-	"encoding/json"
 	"net/http"
 	"net/http/httptest"
 	"path"
 	"testing"
 
 	"github.com/go-kit/kit/log"
-	"github.com/prometheus/prometheus/pkg/rulefmt"
 	"github.com/stretchr/testify/require"
 	"github.com/thanos-io/thanos/pkg/objstore"
 	"github.com/weaveworks/common/user"
 
-	"github.com/cortexproject/cortex/pkg/chunk"
-	"github.com/cortexproject/cortex/pkg/ruler/rulespb"
-	"github.com/cortexproject/cortex/pkg/ruler/rulestore"
-	"github.com/cortexproject/cortex/pkg/ruler/rulestore/objectclient"
 	"github.com/cortexproject/cortex/pkg/storage/tsdb"
 )
 
 func TestDeleteTenant(t *testing.T) {
 	bkt := objstore.NewInMemBucket()
-	api := newTenantDeletionAPI(bkt, nil, nil, log.NewNopLogger())
+	api := newTenantDeletionAPI(bkt, nil, log.NewNopLogger())
 
 	{
 		resp := httptest.NewRecorder()
 		api.DeleteTenant(resp, &http.Request{})
-		require.Equal(t, http.StatusBadRequest, resp.Code)
+		require.Equal(t, http.StatusUnauthorized, resp.Code)
 	}
 
 	{
@@ -86,122 +80,11 @@ func TestDeleteTenantStatus(t *testing.T) {
 				require.NoError(t, bkt.Upload(context.Background(), objName, bytes.NewReader(data)))
 			}
 
-			api := newTenantDeletionAPI(bkt, nil, nil, log.NewNopLogger())
+			api := newTenantDeletionAPI(bkt, nil, log.NewNopLogger())
 
 			res, err := api.isBlocksForUserDeleted(context.Background(), username)
 			require.NoError(t, err)
 			require.Equal(t, tc.expectedBlocksDeleted, res)
 		})
 	}
 }
-
-func TestDeleteTenantRuleGroups(t *testing.T) {
-	ruleGroups := []ruleGroupKey{
-		{user: "userA", namespace: "namespace", group: "group"},
-		{user: "userB", namespace: "namespace1", group: "group"},
-		{user: "userB", namespace: "namespace2", group: "group"},
-	}
-
-	obj, rs := setupRuleGroupsStore(t, ruleGroups)
-	require.Equal(t, 3, obj.GetObjectCount())
-
-	api := newTenantDeletionAPI(objstore.NewInMemBucket(), nil, rs, log.NewNopLogger())
-
-	{
-		callDeleteTenantAPI(t, api, "user-with-no-rule-groups")
-		require.Equal(t, 3, obj.GetObjectCount())
-
-		verifyExpectedDeletedRuleGroupsForUser(t, api, "user-with-no-rule-groups", true) // Has no rule groups
-		verifyExpectedDeletedRuleGroupsForUser(t, api, "userA", false)
-		verifyExpectedDeletedRuleGroupsForUser(t, api, "userB", false)
-	}
-
-	{
-		callDeleteTenantAPI(t, api, "userA")
-		require.Equal(t, 2, obj.GetObjectCount())
-
-		verifyExpectedDeletedRuleGroupsForUser(t, api, "user-with-no-rule-groups", true) // Has no rule groups
-		verifyExpectedDeletedRuleGroupsForUser(t, api, "userA", true)                    // Just deleted.
-		verifyExpectedDeletedRuleGroupsForUser(t, api, "userB", false)
-	}
-
-	{
-		callDeleteTenantAPI(t, api, "userB")
-		require.Equal(t, 0, obj.GetObjectCount())
-
-		verifyExpectedDeletedRuleGroupsForUser(t, api, "user-with-no-rule-groups", true) // Has no rule groups
-		verifyExpectedDeletedRuleGroupsForUser(t, api, "userA", true)                    // Deleted previously
-		verifyExpectedDeletedRuleGroupsForUser(t, api, "userB", true)                    // Just deleted
-	}
-}
-
-func TestDeleteTenantRuleGroupsWithReadOnlyStore(t *testing.T) {
-	ruleGroups := []ruleGroupKey{
-		{user: "userA", namespace: "namespace", group: "group"},
-		{user: "userB", namespace: "namespace1", group: "group"},
-		{user: "userB", namespace: "namespace2", group: "group"},
-	}
-
-	obj, rs := setupRuleGroupsStore(t, ruleGroups)
-	require.Equal(t, 3, obj.GetObjectCount())
-
-	rs = &readOnlyRuleStore{RuleStore: rs}
-
-	api := newTenantDeletionAPI(objstore.NewInMemBucket(), nil, rs, log.NewNopLogger())
-
-	// Make sure there is no error reported.
-	callDeleteTenantAPI(t, api, "userA")
-	require.Equal(t, 3, obj.GetObjectCount())
-
-	verifyExpectedDeletedRuleGroupsForUser(t, api, "userA", false) // Cannot delete from read-only store.
-	verifyExpectedDeletedRuleGroupsForUser(t, api, "userB", false)
-}
-
-func callDeleteTenantAPI(t *testing.T, api *TenantDeletionAPI, userID string) {
-	ctx := user.InjectOrgID(context.Background(), userID)
-
-	req := &http.Request{}
-	resp := httptest.NewRecorder()
-	api.DeleteTenant(resp, req.WithContext(ctx))
-
-	require.Equal(t, http.StatusOK, resp.Code)
-}
-
-func verifyExpectedDeletedRuleGroupsForUser(t *testing.T, api *TenantDeletionAPI, userID string, expected bool) {
-	ctx := user.InjectOrgID(context.Background(), userID)
-
-	req := &http.Request{}
-	resp := httptest.NewRecorder()
-	api.DeleteTenantStatus(resp, req.WithContext(ctx))
-
-	require.Equal(t, http.StatusOK, resp.Code)
-
-	deleteResp := &DeleteTenantStatusResponse{}
-	require.NoError(t, json.Unmarshal(resp.Body.Bytes(), deleteResp))
-	require.Equal(t, expected, deleteResp.RuleGroupsDeleted)
-}
-
-func setupRuleGroupsStore(t *testing.T, ruleGroups []ruleGroupKey) (*chunk.MockStorage, rulestore.RuleStore) {
-	obj := chunk.NewMockStorage()
-	rs := objectclient.NewRuleStore(obj, 5, log.NewNopLogger())
-
-	// "upload" rule groups
-	for _, key := range ruleGroups {
-		desc := rulespb.ToProto(key.user, key.namespace, rulefmt.RuleGroup{Name: key.group})
-		require.NoError(t, rs.SetRuleGroup(context.Background(), key.user, key.namespace, desc))
-	}
-
-	return obj, rs
-}
-
-type ruleGroupKey struct {
-	user, namespace, group string
-}
-
-type readOnlyRuleStore struct {
-	rulestore.RuleStore
-}
-
-func (r *readOnlyRuleStore) SupportsModifications() bool {
-	return false
-}

pkg/cortex/modules.go

@@ -639,13 +639,6 @@ func (t *Cortex) initRulerStorage() (serv services.Service, err error) {
 		return
 	}
 
-	// Purger didn't use ruler storage before, but now it does. However empty configuration just causes error,
-	// so to preserve previous purger behaviour, we simply disable it.
-	if t.Cfg.isModuleEnabled(Purger) && t.Cfg.Ruler.StoreConfig.IsDefaults() {
-		level.Info(util_log.Logger).Log("msg", "Ruler storage is not configured. If you want to use tenant deletion API and delete rule groups, please configure ruler storage.")
-		return
-	}
-
 	if !t.Cfg.Ruler.StoreConfig.IsDefaults() {
 		t.RulerStorage, err = ruler.NewLegacyRuleStore(t.Cfg.Ruler.StoreConfig, rules.FileLoader{}, util_log.Logger)
 	} else {
@@ -810,7 +803,7 @@ func (t *Cortex) initTenantDeletionAPI() (services.Service, error) {
 	}
 
 	// t.RulerStorage can be nil when running in single-binary mode, and rule storage is not configured.
-	tenantDeletionAPI, err := purger.NewTenantDeletionAPI(t.Cfg.BlocksStorage, t.Overrides, t.RulerStorage, util_log.Logger, prometheus.DefaultRegisterer)
+	tenantDeletionAPI, err := purger.NewTenantDeletionAPI(t.Cfg.BlocksStorage, t.Overrides, util_log.Logger, prometheus.DefaultRegisterer)
 	if err != nil {
 		return nil, err
 	}
@@ -895,7 +888,7 @@ func (t *Cortex) setupModuleManager() error {
 		Compactor:                {API, MemberlistKV, Overrides},
 		StoreGateway:             {API, Overrides, MemberlistKV},
 		ChunksPurger:             {Store, DeleteRequestsStore, API},
-		TenantDeletion:           {Store, API, Overrides, RulerStorage},
+		TenantDeletion:           {Store, API, Overrides},
 		Purger:                   {ChunksPurger, TenantDeletion},
 		TenantFederation:         {Queryable},
 		All:                      {QueryFrontend, Querier, Ingester, Distributor, TableManager, Purger, StoreGateway, Ruler},

pkg/ruler/ruler.go

@@ -34,6 +34,7 @@ import (
 	"github.com/cortexproject/cortex/pkg/util"
 	"github.com/cortexproject/cortex/pkg/util/flagext"
 	"github.com/cortexproject/cortex/pkg/util/grpcclient"
+	util_log "github.com/cortexproject/cortex/pkg/util/log"
 	"github.com/cortexproject/cortex/pkg/util/services"
 	"github.com/cortexproject/cortex/pkg/util/validation"
 )
@@ -771,3 +772,24 @@ func (r *Ruler) AssertMaxRulesPerRuleGroup(userID string, rules int) error {
 	}
 	return fmt.Errorf(errMaxRulesPerRuleGroupPerUserLimitExceeded, limit, rules)
 }
+
+func (r *Ruler) DeleteTenantConfiguration(w http.ResponseWriter, req *http.Request) {
+	logger := util_log.WithContext(req.Context(), r.logger)
+
+	userID, err := tenant.TenantID(req.Context())
+	if err != nil {
+		// When Cortex is running, it uses Auth Middleware for checking X-Scope-OrgID and injecting tenant into context.
+		// Auth Middleware sends http.StatusUnauthorized if X-Scope-OrgID is missing, so we do too here, for consistency.
+		http.Error(w, err.Error(), http.StatusUnauthorized)
+		return
+	}
+
+	err = r.store.DeleteNamespace(req.Context(), userID, "") // Empty namespace = delete all rule groups.
+	if err != nil && !errors.Is(err, rulestore.ErrGroupNamespaceNotFound) {
+		respondError(logger, w, err.Error())
+		return
+	}
+
+	level.Info(logger).Log("msg", "deleted all tenant rule groups", "user", userID)
+	w.WriteHeader(http.StatusOK)
+}