feat(NODE-3470): retry selects another mongos

src/operations/execute_operation.ts

@@ -18,6 +18,7 @@ import {
 import type { MongoClient } from '../mongo_client';
 import { ReadPreference } from '../read_preference';
 import type { Server } from '../sdam/server';
+import type { ServerDescription } from '../sdam/server_description';
 import {
   sameServerSelector,
   secondaryWritableServerSelector,
@@ -183,7 +184,8 @@ export async function executeOperation<
       return await retryOperation(operation, operationError, {
         session,
         topology,
-        selector
+        selector,
+        previousServer: server.description
       });
     }
     throw operationError;
@@ -199,6 +201,7 @@ type RetryOptions = {
   session: ClientSession;
   topology: Topology;
   selector: ReadPreference | ServerSelector;
+  previousServer: ServerDescription;
 };
 
 async function retryOperation<
@@ -207,7 +210,7 @@ async function retryOperation<
 >(
   operation: T,
   originalError: MongoError,
-  { session, topology, selector }: RetryOptions
+  { session, topology, selector, previousServer }: RetryOptions
 ): Promise<TResult> {
   const isWriteOperation = operation.hasAspect(Aspect.WRITE_OPERATION);
   const isReadOperation = operation.hasAspect(Aspect.READ_OPERATION);
@@ -243,7 +246,8 @@ async function retryOperation<
   // select a new server, and attempt to retry the operation
   const server = await topology.selectServerAsync(selector, {
     session,
-    operationName: operation.commandName
+    operationName: operation.commandName,
+    previousServer: previousServer
   });
 
   if (isWriteOperation && !supportsRetryableWrites(server)) {

src/sdam/server_selection.ts

@@ -14,7 +14,8 @@ export const MIN_SECONDARY_WRITE_WIRE_VERSION = 13;
 /** @internal */
 export type ServerSelector = (
   topologyDescription: TopologyDescription,
-  servers: ServerDescription[]
+  servers: ServerDescription[],
+  deprioritized?: ServerDescription[]
 ) => ServerDescription[];
 
 /**
@@ -266,7 +267,8 @@ export function readPreferenceServerSelector(readPreference: ReadPreference): Se
 
   return (
     topologyDescription: TopologyDescription,
-    servers: ServerDescription[]
+    servers: ServerDescription[],
+    deprioritized: ServerDescription[] = []
   ): ServerDescription[] => {
     const commonWireVersion = topologyDescription.commonWireVersion;
     if (
@@ -287,13 +289,15 @@ export function readPreferenceServerSelector(readPreference: ReadPreference): Se
       return [];
     }
 
-    if (
-      topologyDescription.type === TopologyType.Single ||
-      topologyDescription.type === TopologyType.Sharded
-    ) {
+    if (topologyDescription.type === TopologyType.Single) {
       return latencyWindowReducer(topologyDescription, servers.filter(knownFilter));
     }
 
+    if (topologyDescription.type === TopologyType.Sharded) {
+      const selectable = servers.length > 0 ? servers : deprioritized;
+      return latencyWindowReducer(topologyDescription, selectable.filter(knownFilter));
+    }
+
     const mode = readPreference.mode;
     if (mode === ReadPreference.PRIMARY) {
       return servers.filter(primaryFilter);

src/sdam/topology.ts

@@ -110,6 +110,7 @@ export interface ServerSelectionRequest {
   timeoutController: TimeoutController;
   operationName: string;
   waitingLogged: boolean;
+  previousServer?: ServerDescription;
 }
 
 /** @internal */
@@ -175,6 +176,7 @@ export interface SelectServerOptions {
   serverSelectionTimeoutMS?: number;
   session?: ClientSession;
   operationName: string;
+  previousServer?: ServerDescription;
 }
 
 /** @public */
@@ -598,7 +600,8 @@ export class Topology extends TypedEventEmitter<TopologyEvents> {
       timeoutController: new TimeoutController(options.serverSelectionTimeoutMS),
       startTime: now(),
       operationName: options.operationName,
-      waitingLogged: false
+      waitingLogged: false,
+      previousServer: options.previousServer
     };
 
     waitQueueMember.timeoutController.signal.addEventListener('abort', () => {
@@ -930,8 +933,13 @@ function processWaitQueue(topology: Topology) {
     let selectedDescriptions;
     try {
       const serverSelector = waitQueueMember.serverSelector;
+      const previousServer = waitQueueMember.previousServer;
       selectedDescriptions = serverSelector
-        ? serverSelector(topology.description, serverDescriptions)
+        ? serverSelector(
+            topology.description,
+            serverDescriptions,
+            previousServer ? [previousServer] : []
+          )
         : serverDescriptions;
     } catch (e) {
       waitQueueMember.timeoutController.clear();

test/integration/server-selection/server_selection.prose.sharded_retryable_reads.test.ts

@@ -0,0 +1,161 @@
+import { expect } from 'chai';
+
+import type { CommandFailedEvent, CommandSucceededEvent } from '../../mongodb';
+
+const TEST_METADATA = { requires: { mongodb: '>=4.2.9', topology: 'sharded' } };
+const FAIL_COMMAND = {
+  configureFailPoint: 'failCommand',
+  mode: { times: 1 },
+  data: {
+    failCommands: ['find'],
+    errorCode: 6,
+    closeConnection: true
+  }
+};
+const DISABLE_FAIL_COMMAND = {
+  configureFailPoint: 'failCommand',
+  mode: 'off',
+  data: {
+    failCommands: ['find'],
+    errorCode: 6,
+    closeConnection: true
+  }
+};
+
+describe('Server Selection Sharded Retryable Reads Prose tests', function () {
+  context('Retryable Reads Are Retried on a Different mongos if One is Available', function () {
+    const commandFailedEvents: CommandFailedEvent[] = [];
+    let client;
+    let utilClientOne;
+    let utilClientTwo;
+
+    // This test MUST be executed against a sharded cluster that has at least two
+    // mongos instances.
+    // 1. Ensure that a test is run against a sharded cluster that has at least two
+    //    mongoses. If there are more than two mongoses in the cluster, pick two to
+    //    test against.
+    beforeEach(async function () {
+      const uri = this.configuration.url({
+        monitorCommands: true,
+        useMultipleMongoses: true
+      });
+
+      // 3. Create a client with ``retryReads=true`` that connects to the cluster,
+      //    providing the two selected mongoses as seeds.
+      client = this.configuration.newClient(uri, {
+        monitorCommands: true,
+        retryReads: true
+      });
+      client.on('commandFailed', event => {
+        commandFailedEvents.push(event);
+      });
+      await client.connect();
+      const seeds = client.topology.s.seedlist.map(address => address.toString());
+
+      // 2. Create a client per mongos using the direct connection, and configure the
+      //    following fail points on each mongos::
+      //      {
+      //          configureFailPoint: "failCommand",
+      //          mode: { times: 1 },
+      //          data: {
+      //              failCommands: ["find"],
+      //              errorCode: 6,
+      //              closeConnection: true
+      //          }
+      //      }
+      utilClientOne = this.configuration.newClient(`mongodb://${seeds[0]}`, {
+        directConnection: true
+      });
+      utilClientTwo = this.configuration.newClient(`mongodb://${seeds[1]}`, {
+        directConnection: true
+      });
+      await utilClientOne.db('admin').command(FAIL_COMMAND);
+      await utilClientTwo.db('admin').command(FAIL_COMMAND);
+    });
+
+    afterEach(async function () {
+      await client?.close();
+      await utilClientOne.db('admin').command(DISABLE_FAIL_COMMAND);
+      await utilClientTwo.db('admin').command(DISABLE_FAIL_COMMAND);
+      await utilClientOne?.close();
+      await utilClientTwo?.close();
+    });
+
+    // 4. Enable command monitoring, and execute a ``find`` command that is
+    //    supposed to fail on both mongoses.
+    // 5. Asserts that there were failed command events from each mongos.
+    // 6. Disable the fail points.
+    it('retries on a different mongos', TEST_METADATA, async function () {
+      await client
+        .db('test')
+        .collection('test')
+        .find()
+        .toArray()
+        .catch(() => null);
+      expect(commandFailedEvents[0].address).to.not.equal(commandFailedEvents[1].address);
+    });
+  });
+
+  // 1. Ensure that a test is run against a sharded cluster. If there are multiple
+  // mongoses in the cluster, pick one to test against.
+  context('Retryable Reads Are Retried on the Same mongos if No Others are Available', function () {
+    const commandFailedEvents: CommandFailedEvent[] = [];
+    const commandSucceededEvents: CommandSucceededEvent[] = [];
+    let client;
+    let utilClient;
+
+    beforeEach(async function () {
+      const uri = this.configuration.url({
+        monitorCommands: true
+      });
+
+      // 3. Create a client with ``retryReads=true`` that connects to the cluster,
+      //     providing the selected mongos as the seed.
+      client = this.configuration.newClient(uri, {
+        monitorCommands: true,
+        retryReads: true
+      });
+      client.on('commandFailed', event => {
+        commandFailedEvents.push(event);
+      });
+      client.on('commandSucceeded', event => {
+        commandSucceededEvents.push(event);
+      });
+
+      // 2. Create a client that connects to the mongos using the direct connection,
+      //     and configure the following fail point on the mongos::
+      //       {
+      //           configureFailPoint: "failCommand",
+      //           mode: { times: 1 },
+      //           data: {
+      //               failCommands: ["find"],
+      //               errorCode: 6,
+      //               closeConnection: true
+      //           }
+      //       }
+      utilClient = this.configuration.newClient(uri, {
+        directConnection: true
+      });
+      await utilClient.db('admin').command(FAIL_COMMAND);
+    });
+
+    afterEach(async function () {
+      await client?.close();
+      await utilClient?.db('admin').command(DISABLE_FAIL_COMMAND);
+      await utilClient?.close();
+    });
+
+    // 4. Enable command monitoring, and execute a ``find`` command.
+    // 5. Asserts that there was a failed command and a successful command event.
+    // 6. Disable the fail point.
+    it('retries on the same mongos', TEST_METADATA, async function () {
+      await client
+        .db('test')
+        .collection('test')
+        .find()
+        .toArray()
+        .catch(() => null);
+      expect(commandFailedEvents[0].address).to.equal(commandSucceededEvents[0].address);
+    });
+  });
+});