SERVER-110919 Ban hint on '_id_' for timeseries collections (#41961)

GitOrigin-RevId: 4d73214d504aa32e91911ae0fe40fbe87af9a039

buildscripts/smoke_tests/server_collection_write_path.yml

@@ -44,6 +44,7 @@ suites:
     - jstests/core/timeseries/write/timeseries_delete_with_meta_concurrent.js
     - jstests/core/timeseries/write/timeseries_findAndModify_deletes_hints.js
     - jstests/core/timeseries/write/timeseries_findAndModify_updates_hints.js
+    - jstests/core/timeseries/write/timeseries_hint_id.js
     - jstests/core/timeseries/write/timeseries_insert.js
     - jstests/core/timeseries/write/timeseries_insert_after_delete.js
     - jstests/core/timeseries/write/timeseries_insert_after_update.js

buildscripts/smoke_tests/storage_execution.yml

@@ -45,6 +45,7 @@ suites:
     - jstests/core/timeseries/write/timeseries_delete_with_meta_concurrent.js
     - jstests/core/timeseries/write/timeseries_findAndModify_deletes_hints.js
     - jstests/core/timeseries/write/timeseries_findAndModify_updates_hints.js
+    - jstests/core/timeseries/write/timeseries_hint_id.js
     - jstests/core/timeseries/write/timeseries_insert.js
     - jstests/core/timeseries/write/timeseries_insert_after_delete.js
     - jstests/core/timeseries/write/timeseries_insert_after_update.js

jstests/core/timeseries/write/timeseries_hint_id.js

@@ -0,0 +1,131 @@
+/**
+ * Tests running the find and delete commands with a hint on the _id index on a timeseries
+ * collection. Verifies that commands specifying a hint on the _id index via "_id_" always fail.
+ * @tags: [
+ *   # This is a backwards-breaking change.
+ *   requires_fcv_83,
+ *   does_not_support_stepdowns,
+ *   # Retryable deletes, for example, not supported on timeseries collections
+ *   does_not_support_retryable_writes,
+ *   requires_non_retryable_commands,
+ *   requires_non_retryable_writes,
+ *   # We need a timeseries collection.
+ *   requires_timeseries,
+ * ]
+ */
+const timeFieldName = "time";
+const metaFieldName = "tag";
+const collName = jsTestName();
+const dbName = jsTestName();
+
+const testDB = db.getSiblingDB(dbName);
+const coll = testDB.getCollection(collName);
+
+assert.commandWorked(
+    testDB.createCollection(coll.getName(), {timeseries: {timeField: timeFieldName, metaField: metaFieldName}}),
+);
+
+const objA = {
+    [timeFieldName]: ISODate(),
+    "measurement": {"A": "cpu"},
+    [metaFieldName]: {a: "A"},
+};
+const objB = {
+    [timeFieldName]: ISODate(),
+    "measurement": {"A": "cpu"},
+    [metaFieldName]: {b: "B"},
+};
+const objC = {
+    [timeFieldName]: ISODate(),
+    "measurement": {"A": "cpu"},
+    [metaFieldName]: {c: "C"},
+};
+
+assert.commandWorked(coll.insert([objA, objB, objC]));
+
+function runCommandOnTSCollection(commandWithHintObjectValue, commandWithHintStringValue) {
+    // No user-facing index exists on the _id field. It only exists on the underlying buckets
+    // collection.
+    assert.commandFailedWithCode(testDB.runCommand(commandWithHintObjectValue), ErrorCodes.BadValue);
+
+    assert.commandFailedWithCode(testDB.runCommand(commandWithHintStringValue), ErrorCodes.BadValue);
+
+    // Create an index on the _id field of the user-facing view collection.
+    assert.commandWorked(coll.createIndex({_id: 1}));
+
+    // This hint will be rewritten as the key pattern (ex:
+    // "hint":{"control.min._id":1,"control.max._id":1}) for the buckets collection.
+    assert.commandWorked(testDB.runCommand(commandWithHintObjectValue));
+
+    assert.commandFailedWithCode(testDB.runCommand(commandWithHintStringValue), ErrorCodes.BadValue);
+
+    // Drop the index on the _id field of the user-facing view collection.
+    assert.commandWorked(coll.dropIndex({_id: 1}));
+}
+
+(function runFindCommand() {
+    runCommandOnTSCollection({find: collName, hint: {_id: 1}}, {find: collName, hint: "_id_"});
+})();
+
+(function runDeleteCommand() {
+    runCommandOnTSCollection(
+        {delete: collName, deletes: [{q: {metaFieldName: "b"}, limit: 0, hint: {_id: 1}}]},
+        {delete: collName, deletes: [{q: {metaFieldName: "c"}, limit: 0, hint: "_id_"}]},
+    );
+    assert.commandWorked(coll.insert([objB, objC]));
+})();
+
+(function runAggregateCommand() {
+    assert.commandWorked(coll.insert([objA]));
+    const matchPipeline = [{$match: {metaFieldName: "a"}}, {$sort: {_id: 1}}, {$project: {_id: 0, time: 0}}];
+    runCommandOnTSCollection(
+        {
+            aggregate: collName,
+            pipeline: matchPipeline,
+            cursor: {},
+            allowDiskUse: true,
+            hint: {_id: 1},
+        },
+        {
+            aggregate: collName,
+            pipeline: matchPipeline,
+            cursor: {},
+            allowDiskUse: true,
+            hint: "_id_",
+        },
+    );
+})();
+
+(function runUpdateCommand() {
+    runCommandOnTSCollection(
+        {
+            update: collName,
+            updates: [
+                {
+                    q: {tag: "b"},
+                    u: {$set: {tag: {b: "RAM"}}},
+                    hint: {_id: 1},
+                    multi: true,
+                },
+            ],
+        },
+        {
+            update: collName,
+            updates: [
+                {
+                    q: {tag: "c"},
+                    u: {$set: {tag: {c: "L1"}}},
+                    hint: "_id_",
+                    multi: true,
+                },
+            ],
+        },
+    );
+})();
+
+(function runCountCommand() {
+    runCommandOnTSCollection(
+        {count: collName, query: {metaFieldName: "a"}, hint: {_id: 1}},
+        {count: collName, query: {metaFieldName: "a"}, hint: "_id_"},
+    );
+})();

src/mongo/db/query/query_planner.cpp

@@ -44,6 +44,7 @@
 #include "mongo/bson/bsonobj.h"
 #include "mongo/bson/bsonobjbuilder.h"
 #include "mongo/bson/bsontypes.h"
+#include "mongo/db/index/index_constants.h"
 #include "mongo/db/index_names.h"
 #include "mongo/db/local_catalog/clustered_collection_options_gen.h"
 #include "mongo/db/matcher/expression_algo.h"
@@ -978,9 +979,18 @@ StatusWith<std::vector<std::unique_ptr<QuerySolution>>> QueryPlanner::plan(
     boost::optional<BSONObj> hintedIndexBson = boost::none;
     if (!params.indexFiltersApplied && !params.querySettingsApplied) {
         if (auto hintObj = query.getFindCommandRequest().getHint(); !hintObj.isEmpty()) {
+            if (params.mainCollectionInfo.stats.isTimeseries &&
+                hintObj.firstElement().valueStringDataSafe() == IndexConstants::kIdIndexName) {
+                // The index name '_id_' is reserved for the _id index on the underlying buckets
+                // collection of a timeseries collection. The user will never be able to specify the
+                // _id index name '_id_' in the hint.
+                return Status(ErrorCodes::BadValue,
+                              "cannot hint on '_id_' for timeseries collections");
+            } else {
                 hintedIndexBson = hintObj;
             }
         }
+    }
 
     // geoNear and text queries *require* an index.
     // Also, if a hint is specified it indicates that we MUST use it.