[ResponseOps][Alerting] Decouple feature IDs from consumers (elastic#183756)

## Summary

This PR aims to decouple the feature IDs from the `consumer` attribute
of rules and alerts.

Towards: elastic#187202
Fixes: elastic#181559
Fixes: elastic#182435

> [!NOTE]  
> Unfortunately, I could not break the PR into smaller pieces. The APIs
could not work anymore with feature IDs and had to convert them to use
rule type IDs. Also, I took the chance and refactored crucial parts of
the authorization class that in turn affected a lot of files. Most of
the changes in the files are minimal and easy to review. The crucial
changes are in the authorization class and some alerting APIs.

## Architecture

### Alerting RBAC model

The Kibana security uses Elasticsearch's [application
privileges](https://www.elastic.co/guide/en/elasticsearch/reference/current/security-api-put-privileges.html#security-api-put-privileges).
This way Kibana can represent and store its privilege models within
Elasticsearch roles. To do that, Kibana security creates actions that
are granted by a specific privilege. Alerting uses its own RBAC model
and is built on top of the existing Kibana security model. The Alerting
RBAC uses the `rule_type_id` and `consumer` attributes to define who
owns the rule and the alerts procured by the rule. To connect the
`rule_type_id` and `consumer` with the Kibana security actions the
Alerting RBAC registers its custom actions. They are constructed as
`alerting:<rule-type-id>/<feature-id>/<alerting-entity>/<operation>`.
Because to authorizate a resource an action has to be generated and
because the action needs a valid feature ID the value of the `consumer`
should be a valid feature ID. For example, the
`alerting:siem.esqlRule/siem/rule/get` action, means that a user with a
role that grants this action can get a rule of type `siem.esqlRule` with
consumer `siem`.

### Problem statement

At the moment the `consumer` attribute should be a valid feature ID.
Though this approach worked well so far it has its limitation.
Specifically:

- Rule types cannot support more than one consumer.
- To associate old rules with a new feature ID required a migration on
the rule's SOs and the alerts documents.
- The API calls are feature ID-oriented and not rule-type-oriented.
- The framework has to be aware of the values of the `consumer`
attribute.
- Feature IDs are tightly coupled with the alerting indices leading to
[bugs](elastic#179082).
- Legacy consumers that are not a valid feature anymore can cause
[bugs](elastic#184595).
- The framework has to be aware of legacy consumers to handle edge
cases.
- The framework has to be aware of specific consumers to handle edge
cases.

### Proposed solution

This PR aims to decouple the feature IDs from consumers. It achieves
that a) by changing the way solutions configure the alerting privileges
when registering a feature and b) by changing the alerting actions. The
schema changes as:

```
// Old formatting
id: 'siem', <--- feature ID
alerting:['siem.queryRule']

// New formatting
id: 'siem', <--- feature ID
alerting: [{ ruleTypeId: 'siem.queryRule', consumers: ['siem'] }] <-- consumer same as the feature ID in the old formatting
```

The new actions are constructed as
`alerting:<rule-type-id>/<consumer>/<alerting-entity>/<operation>`. For
example `alerting:rule-type-id/my-consumer/rule/get`. The new action
means that a user with a role that grants this action can get a rule of
type `rule-type` with consumer `my-consumer`. Changing the action
strings is not considered a breaking change as long as the user's
permission works as before. In our case, this is true because the
consumer will be the same as before (feature ID), and the alerting
security actions will be the same. For example:

**Old formatting**

Schema:
```
id: 'logs', <--- feature ID
alerting:['.es-query'] <-- rule type ID
```

Generated action:

```
alerting:.es-query/logs/rule/get
```

**New formatting**

Schema:
```
id: 'siem', <--- feature ID
alerting: [{ ruleTypeId: '.es-query', consumers: ['logs'] }] <-- consumer same as the feature ID in the old formatting
```

Generated action:

```
alerting:.es-query/logs/rule/get <--- consumer is set as logs and the action is the same as before
```

In both formating the actions are the same thus breaking changes are
avoided.

### Alerting authorization class
The alerting plugin uses and exports the alerting authorization class
(`AlertingAuthorization`). The class is responsible for handling all
authorization actions related to rules and alerts. The class changed to
handle the new actions as described in the above sections. A lot of
methods were renamed, removed, and cleaned up, all method arguments
converted to be an object, and the response signature of some methods
changed. These changes affected various pieces of the code. The changes
in this class are the most important in this PR especially the
`_getAuthorizedRuleTypesWithAuthorizedConsumers` method which is the
cornerstone of the alerting RBAC. Please review carefully.

### Instantiation of the alerting authorization class
The `AlertingAuthorizationClientFactory` is used to create instances of
the `AlertingAuthorization` class. The `AlertingAuthorization` class
needs to perform async operations upon instantiation. Because JS, at the
moment, does not support async instantiation of classes the
`AlertingAuthorization` class was assigning `Promise` objects to
variables that could be resolved later in other phases of the lifecycle
of the class. To improve readability and make the lifecycle of the class
clearer, I separated the construction of the class (initialization) from
the bootstrap process. As a result, getting the `AlertingAuthorization`
class or any client that depends on it (`getRulesClient` for example) is
an async operation.

### Filtering
A lot of routes use the authorization class to get the authorization
filter (`getFindAuthorizationFilter`), a filter that, if applied,
returns only the rule types and consumers the user is authorized to. The
method that returns the filter was built in a way to also support
filtering on top of the authorization filter thus coupling the
authorized filter with router filtering. I believe these two operations
should be decoupled and the filter method should return a filter that
gives you all the authorized rule types. It is the responsibility of the
consumer, router in our case, to apply extra filters on top of the
authorization filter. For that reason, I made all the necessary changes
to decouple them.

### Legacy consumers & producer
A lot of rules and alerts have been created and are still being created
from observability with the `alerts` consumer. When the Alerting RBAC
encounters a rule or alert with `alerts` as a consumer it falls back to
the `producer` of the rule type ID to construct the actions. For example
if a rule with `ruleTypeId: .es-query` and `consumer: alerts` the
alerting action will be constructed as
`alerting:.es-query/stackAlerts/rule/get` where `stackRules` is the
producer of the `.es-query` rule type. The `producer` is used to be used
in alerting authorization but due to its complexity, it was deprecated
and only used as a fallback for the `alerts` consumer. To avoid breaking
changes all feature privileges that specify access to rule types add the
`alerts` consumer when configuring their alerting privileges. By moving
the `alerts` consumer to the registration of the feature we can stop
relying on the `producer`. The `producer` is not used anymore in the
authorization class. In the next PRs the `producer` will removed
entirely.

### Routes
The following changes were introduced to the alerting routes:

- All related routes changed to be rule-type oriented and not feature ID
oriented.
- All related routes support the `ruleTypeIds` and the `consumers`
parameters for filtering. In all routes, the filters are constructed as
`ruleTypeIds: ['foo'] AND consumers: ['bar'] AND authorizationFilter`.
Filtering by consumers is important. In o11y for example, we do not want
to show ES rule types with the `stackAlerts` consumer even if the user
has access to them.
- The `/internal/rac/alerts/_feature_ids` route got deleted as it was
not used anywhere in the codebase and it was internal.

All the changes in the routes are related to internal routes and no
breaking changes are introduced.

### Constants
I moved the o11y and stack rule type IDs to `kbn-rule-data-utils` and
exported all security solution rule type IDs from
`kbn-securitysolution-rules`. I am not a fan of having a centralized
place for the rule type IDs. Ideally, consumers of the framework should
specify keywords like `observablility` (category or subcategory) or even
`apm.*` and the framework should know which rule type IDs to pick up. I
think it is out of the scope of the PR, and at the moment it seems the
most straightforward way to move forward. I will try to clean up as much
as possible in further iterations. If you are interested in the upcoming
work follow this issue elastic#187202.

### Other notable code changes
- Change all instances of feature IDs to rule type IDs.
- `isSiemRuleType`: This is a temporary helper function that is needed
in places where we handle edge cases related to security solution rule
types. Ideally, the framework should be agnostic to the rule types or
consumers. The plan is to be removed entirely in further iterations.
- Rename alerting `PluginSetupContract` and `PluginStartContract` to
`AlertingServerSetup` and `AlertingServerStart`. This made me touch a
lot of files but I could not resist.
- `filter_consumers` was mistakenly exposed to a public API. It was
undocumented.
- Files or functions that were not used anywhere in the codebase got
deleted.
- Change the returned type of the `list` method of the
`RuleTypeRegistry` from `Set<RegistryRuleType>` to `Map<string,
RegistryRuleType>`.
- Assertion of `KueryNode` in tests changed to an assertion of KQL using
`toKqlExpression`.
- Removal of `useRuleAADFields` as it is not used anywhere.

## Testing

> [!CAUTION]
> It is very important to test all the areas of the application where
rules or alerts are being used directly or indirectly. Scenarios to
consider:
> - The correct rules, alerts, and aggregations on top of them are being
shown as expected as a superuser.
> - The correct rules, alerts, and aggregations on top of them are being
shown as expected by a user with limited access to certain features.
> - The changes in this PR are backward compatible with the previous
users' permissions.

### Solutions
Please test and verify that:
- All the rule types you own with all possible combinations of
permissions both in ESS and in Serverless.
- The consumers and rule types make sense when registering the features.
- The consumers and rule types that are passed to the components are the
intended ones.

### ResponseOps
The most important changes are in the alerting authorization class, the
search strategy, and the routes. Please test:
- The rules we own with all possible combinations of permissions.
- The stack alerts page and its solution filtering.
- The categories filtering in the maintenance window UI.

## Risks
> [!WARNING]
> The risks involved in this PR are related to privileges. Specifically:
> - Users with no privileges can access rules and alerts they do not
have access to.
> - Users with privileges cannot access rules and alerts they have
access to.
>
> An excessive list of integration tests is in place to ensure that the
above scenarios will not occur. In the case of a bug, we could a)
release an energy release for serverless and b) backport the fix in ESS.
Given that this PR is intended to be merged in 8.17 we have plenty of
time to test and to minimize the chances of risks.

## FQA

- I noticed that a lot of routes support the `filter` parameter where we
can pass an arbitrary KQL filter. Why we do not use this to filter by
the rule type IDs and the consumers and instead we introduce new
dedicated parameters?

The `filter` parameter should not be exposed in the first place. It
assumes that the consumer of the API knows the underlying structure and
implementation details of the persisted storage API (SavedObject client
API). For example, a valid filter would be
`alerting.attributes.rule_type_id`. In this filter the consumer should
know a) the name of the SO b) the keyword `attributes` (storage
implementation detail) and c) the name of the attribute as it is
persisted in ES (snake case instead of camel case as it is returned by
the APIs). As there is no abstraction layer between the SO and the API,
it makes it very difficult to make changes in the persistent schema or
the APIs. For all the above I decided to introduce new query parameters
where the alerting framework has total control over it.

- I noticed in the code a lot of instances where the consumer is used.
Should not remove any logic around consumers?

This PR is a step forward making the framework as agnostic as possible.
I had to keep the scope of the PR as contained as possible. We will get
there. It needs time :).

- I noticed a lot of hacks like checking if the rule type is `siem`.
Should not remove the hacks?

This PR is a step forward making the framework as agnostic as possible.
I had to keep the scope of the PR as contained as possible. We will get
there. It needs time :).

- I hate the "Role visibility" dropdown. Can we remove it?

I also do not like it. The goal is to remove it. Follow
elastic#189997.

---------

Co-authored-by: kibanamachine <[email protected]>
Co-authored-by: Aleh Zasypkin <[email protected]>
Co-authored-by: Paula Borgonovi <[email protected]>

Author: 4 people

packages/kbn-alerting-types/search_strategy_types.ts

@@ -8,7 +8,6 @@
  */
 
 import type { IEsSearchRequest, IEsSearchResponse } from '@kbn/search-types';
-import type { ValidFeatureId } from '@kbn/rule-data-utils';
 import type {
   MappingRuntimeFields,
   QueryDslFieldAndFormat,
@@ -18,7 +17,8 @@ import type {
 import type { Alert } from './alert_type';
 
 export type RuleRegistrySearchRequest = IEsSearchRequest & {
-  featureIds: ValidFeatureId[];
+  ruleTypeIds: string[];
+  consumers?: string[];
   fields?: QueryDslFieldAndFormat[];
   query?: Pick<QueryDslQueryContainer, 'bool' | 'ids'>;
   sort?: SortCombinations[];

packages/kbn-alerts-as-data-utils/src/schemas/generated/observability_threshold_schema.ts

@@ -0,0 +1,90 @@
+/*
+ * Copyright Elasticsearch B.V. and/or licensed to Elasticsearch B.V. under one
+ * or more contributor license agreements. Licensed under the "Elastic License
+ * 2.0", the "GNU Affero General Public License v3.0 only", and the "Server Side
+ * Public License v 1"; you may not use this file except in compliance with, at
+ * your election, the "Elastic License 2.0", the "GNU Affero General Public
+ * License v3.0 only", or the "Server Side Public License, v 1".
+ */
+// ---------------------------------- WARNING ----------------------------------
+// this file was generated, and should not be edited by hand
+// ---------------------------------- WARNING ----------------------------------
+import * as rt from 'io-ts';
+import { Either } from 'fp-ts/lib/Either';
+import { AlertSchema } from './alert_schema';
+import { EcsSchema } from './ecs_schema';
+const ISO_DATE_PATTERN = /^d{4}-d{2}-d{2}Td{2}:d{2}:d{2}.d{3}Z$/;
+export const IsoDateString = new rt.Type<string, string, unknown>(
+  'IsoDateString',
+  rt.string.is,
+  (input, context): Either<rt.Errors, string> => {
+    if (typeof input === 'string' && ISO_DATE_PATTERN.test(input)) {
+      return rt.success(input);
+    } else {
+      return rt.failure(input, context);
+    }
+  },
+  rt.identity
+);
+export type IsoDateStringC = typeof IsoDateString;
+export const schemaUnknown = rt.unknown;
+export const schemaUnknownArray = rt.array(rt.unknown);
+export const schemaString = rt.string;
+export const schemaStringArray = rt.array(schemaString);
+export const schemaNumber = rt.number;
+export const schemaNumberArray = rt.array(schemaNumber);
+export const schemaDate = rt.union([IsoDateString, schemaNumber]);
+export const schemaDateArray = rt.array(schemaDate);
+export const schemaDateRange = rt.partial({
+  gte: schemaDate,
+  lte: schemaDate,
+});
+export const schemaDateRangeArray = rt.array(schemaDateRange);
+export const schemaStringOrNumber = rt.union([schemaString, schemaNumber]);
+export const schemaStringOrNumberArray = rt.array(schemaStringOrNumber);
+export const schemaBoolean = rt.boolean;
+export const schemaBooleanArray = rt.array(schemaBoolean);
+const schemaGeoPointCoords = rt.type({
+  type: schemaString,
+  coordinates: schemaNumberArray,
+});
+const schemaGeoPointString = schemaString;
+const schemaGeoPointLatLon = rt.type({
+  lat: schemaNumber,
+  lon: schemaNumber,
+});
+const schemaGeoPointLocation = rt.type({
+  location: schemaNumberArray,
+});
+const schemaGeoPointLocationString = rt.type({
+  location: schemaString,
+});
+export const schemaGeoPoint = rt.union([
+  schemaGeoPointCoords,
+  schemaGeoPointString,
+  schemaGeoPointLatLon,
+  schemaGeoPointLocation,
+  schemaGeoPointLocationString,
+]);
+export const schemaGeoPointArray = rt.array(schemaGeoPoint);
+// prettier-ignore
+const ObservabilityThresholdAlertRequired = rt.type({
+});
+// prettier-ignore
+const ObservabilityThresholdAlertOptional = rt.partial({
+  'kibana.alert.context': schemaUnknown,
+  'kibana.alert.evaluation.threshold': schemaStringOrNumber,
+  'kibana.alert.evaluation.value': schemaStringOrNumber,
+  'kibana.alert.evaluation.values': schemaStringOrNumberArray,
+  'kibana.alert.group': rt.array(
+    rt.partial({
+      field: schemaStringArray,
+      value: schemaStringArray,
+    })
+  ),
+});
+
+// prettier-ignore
+export const ObservabilityThresholdAlertSchema = rt.intersection([ObservabilityThresholdAlertRequired, ObservabilityThresholdAlertOptional, AlertSchema, EcsSchema]);
+// prettier-ignore
+export type ObservabilityThresholdAlert = rt.TypeOf<typeof ObservabilityThresholdAlertSchema>;

packages/kbn-alerts-grouping/src/components/alerts_grouping.test.tsx

@@ -23,7 +23,8 @@ import { groupingSearchResponse } from '../mocks/grouping_query.mock';
 import { useAlertsGroupingState } from '../contexts/alerts_grouping_context';
 import { I18nProvider } from '@kbn/i18n-react';
 import {
-  mockFeatureIds,
+  mockRuleTypeIds,
+  mockConsumers,
   mockDate,
   mockGroupingProps,
   mockGroupingId,
@@ -146,7 +147,8 @@ describe('AlertsGrouping', () => {
       expect.objectContaining({
         params: {
           aggregations: {},
-          featureIds: mockFeatureIds,
+          ruleTypeIds: mockRuleTypeIds,
+          consumers: mockConsumers,
           groupByField: 'kibana.alert.rule.name',
           filters: [
             {

packages/kbn-alerts-grouping/src/components/alerts_grouping.tsx

@@ -66,7 +66,7 @@ const AlertsGroupingInternal = <T extends BaseAlertsGroupAggregations>(
   const {
     groupingId,
     services,
-    featureIds,
+    ruleTypeIds,
     defaultGroupingOptions,
     defaultFilters,
     globalFilters,
@@ -79,7 +79,7 @@ const AlertsGroupingInternal = <T extends BaseAlertsGroupAggregations>(
   const { grouping, updateGrouping } = useAlertsGroupingState(groupingId);
 
   const { dataView } = useAlertsDataView({
-    featureIds,
+    ruleTypeIds,
     dataViewsService: dataViews,
     http,
     toasts: notifications.toasts,
@@ -252,7 +252,7 @@ const typedMemo: <T>(c: T) => T = memo;
  *
  * return (
  *   <AlertsGrouping<YourAggregationsType>
- *     featureIds={[...]}
+ *     ruleTypeIds={[...]}
  *     globalQuery={{ query: ..., language: 'kql' }}
  *     globalFilters={...}
  *     from={...}

packages/kbn-alerts-grouping/src/components/alerts_grouping_level.test.tsx

@@ -55,6 +55,10 @@ const mockGroupingLevelProps: Omit<AlertsGroupingLevelProps, 'children'> = {
 describe('AlertsGroupingLevel', () => {
   let buildEsQuerySpy: jest.SpyInstance;
 
+  beforeEach(() => {
+    jest.clearAllMocks();
+  });
+
   beforeAll(() => {
     buildEsQuerySpy = jest.spyOn(buildEsQueryModule, 'buildEsQuery');
   });
@@ -119,4 +123,58 @@ describe('AlertsGroupingLevel', () => {
       Object.keys(groupingSearchResponse.aggregations)
     );
   });
+
+  it('should calls useGetAlertsGroupAggregationsQuery with correct props', () => {
+    render(
+      <AlertsGroupingLevel {...mockGroupingLevelProps}>
+        {() => <span data-test-subj="grouping-level" />}
+      </AlertsGroupingLevel>
+    );
+
+    expect(mockUseGetAlertsGroupAggregationsQuery.mock.calls).toMatchInlineSnapshot(`
+      Array [
+        Array [
+          Object {
+            "enabled": true,
+            "http": Object {
+              "get": [MockFunction],
+            },
+            "params": Object {
+              "aggregations": Object {},
+              "consumers": Array [
+                "stackAlerts",
+              ],
+              "filters": Array [
+                Object {
+                  "bool": Object {
+                    "filter": Array [],
+                    "must": Array [],
+                    "must_not": Array [],
+                    "should": Array [],
+                  },
+                },
+                Object {
+                  "range": Object {
+                    "kibana.alert.time_range": Object {
+                      "gte": "2020-07-07T08:20:18.966Z",
+                      "lte": "2020-07-08T08:20:18.966Z",
+                    },
+                  },
+                },
+              ],
+              "groupByField": "selectedGroup",
+              "pageIndex": 0,
+              "pageSize": 10,
+              "ruleTypeIds": Array [
+                ".es-query",
+              ],
+            },
+            "toasts": Object {
+              "addDanger": [MockFunction],
+            },
+          },
+        ],
+      ]
+    `);
+  });
 });

packages/kbn-alerts-grouping/src/components/alerts_grouping_level.tsx

@@ -46,7 +46,8 @@ const DEFAULT_FILTERS: Filter[] = [];
 const typedMemo: <T>(c: T) => T = memo;
 export const AlertsGroupingLevel = typedMemo(
   <T extends BaseAlertsGroupAggregations>({
-    featureIds,
+    ruleTypeIds,
+    consumers,
     defaultFilters = DEFAULT_FILTERS,
     from,
     getGrouping,
@@ -86,7 +87,8 @@ export const AlertsGroupingLevel = typedMemo(
 
     const aggregationsQuery = useMemo<UseGetAlertsGroupAggregationsQueryProps['params']>(() => {
       return {
-        featureIds,
+        ruleTypeIds,
+        consumers,
         groupByField: selectedGroup,
         aggregations: getAggregationsByGroupingField(selectedGroup)?.reduce(
           (acc, val) => Object.assign(acc, val),
@@ -107,12 +109,13 @@ export const AlertsGroupingLevel = typedMemo(
         pageSize,
       };
     }, [
-      featureIds,
+      consumers,
       filters,
       from,
       getAggregationsByGroupingField,
       pageIndex,
       pageSize,
+      ruleTypeIds,
       selectedGroup,
       to,
     ]);

packages/kbn-alerts-grouping/src/mocks/grouping_props.mock.tsx

@@ -8,12 +8,12 @@
  */
 
 import React from 'react';
-import { AlertConsumers } from '@kbn/rule-data-utils';
 import { AlertsGroupingProps } from '../types';
 
 export const mockGroupingId = 'test';
 
-export const mockFeatureIds = [AlertConsumers.STACK_ALERTS];
+export const mockRuleTypeIds = ['.es-query'];
+export const mockConsumers = ['stackAlerts'];
 
 export const mockDate = {
   from: '2020-07-07T08:20:18.966Z',
@@ -30,7 +30,8 @@ export const mockOptions = [
 export const mockGroupingProps: Omit<AlertsGroupingProps, 'children'> = {
   ...mockDate,
   groupingId: mockGroupingId,
-  featureIds: mockFeatureIds,
+  ruleTypeIds: mockRuleTypeIds,
+  consumers: mockConsumers,
   defaultGroupingOptions: mockOptions,
   getAggregationsByGroupingField: () => [],
   getGroupStats: () => [{ title: 'Stat', component: <span /> }],

packages/kbn-alerts-grouping/src/mocks/grouping_query.mock.ts

@@ -11,12 +11,12 @@ export const getQuery = ({
   selectedGroup,
   uniqueValue,
   timeRange,
-  featureIds,
+  ruleTypeIds,
 }: {
   selectedGroup: string;
   uniqueValue: string;
   timeRange: { from: string; to: string };
-  featureIds: string[];
+  ruleTypeIds: string[];
 }) => ({
   _source: false,
   aggs: {
@@ -52,7 +52,7 @@ export const getQuery = ({
       },
     },
   },
-  feature_ids: featureIds,
+  rule_type_ids: ruleTypeIds,
   query: {
     bool: {
       filter: [

packages/kbn-alerts-grouping/src/types.ts

@@ -8,7 +8,6 @@
  */
 
 import type { Filter, Query } from '@kbn/es-query';
-import { ValidFeatureId } from '@kbn/rule-data-utils';
 import type { NotificationsStart } from '@kbn/core-notifications-browser';
 import type { DataViewsServicePublic } from '@kbn/data-views-plugin/public/types';
 import type { HttpSetup } from '@kbn/core-http-browser';
@@ -63,9 +62,13 @@ export interface AlertsGroupingProps<
    */
   defaultGroupingOptions: GroupOption[];
   /**
-   * The alerting feature ids this grouping covers
+   * The alerting rule type ids this grouping covers
    */
-  featureIds: ValidFeatureId[];
+  ruleTypeIds: string[];
+  /**
+   * The alerting consumers this grouping covers
+   */
+  consumers?: string[];
   /**
    * Time filter start
    */

packages/kbn-alerts-ui-shared/src/action_variables/transforms.test.ts

@@ -289,5 +289,6 @@ function getAlertType(actionVariables: ActionVariables): RuleType {
     producer: ALERTING_FEATURE_ID,
     minimumLicenseRequired: 'basic',
     enabledInLicense: true,
+    category: 'my-category',
   };
 }