SERVER-24600 Increase interruptibility of RemoteCommandTargeter::findHost.

By making more calls of RemoteCommandTargeter::findHost interruptible, this
change speeds up the shutdown of mongos when no config servers are discoverable.

8 files changed, 14 insertions, 15 deletions

diff --git a/src/mongo/s/balancer/migration_manager.cpp b/src/mongo/s/balancer/migration_manager.cpp
index b73199ec848..ba4a488ffcb 100644
--- a/src/mongo/s/balancer/migration_manager.cpp
+++ b/src/mongo/s/balancer/migration_manager.cpp
@@ -327,8 +327,7 @@ shared_ptr<Notification<Status>> MigrationManager::_schedule(
 
     const auto fromShard = fromShardStatus.getValue();
     auto fromHostStatus =
-        fromShard->getTargeter()->findHost(ReadPreferenceSetting{ReadPreference::PrimaryOnly},
-                                           RemoteCommandTargeter::selectFindHostMaxWaitTime(txn));
+        fromShard->getTargeter()->findHost(txn, ReadPreferenceSetting{ReadPreference::PrimaryOnly});
     if (!fromHostStatus.isOK()) {
         return std::make_shared<Notification<Status>>(std::move(fromHostStatus.getStatus()));
     }
diff --git a/src/mongo/s/catalog/replset/sharding_catalog_manager_impl.cpp b/src/mongo/s/catalog/replset/sharding_catalog_manager_impl.cpp
index f58c323fd12..88c7d4c482f 100644
--- a/src/mongo/s/catalog/replset/sharding_catalog_manager_impl.cpp
+++ b/src/mongo/s/catalog/replset/sharding_catalog_manager_impl.cpp
@@ -325,8 +325,7 @@ StatusWith<Shard::CommandResponse> ShardingCatalogManagerImpl::_runCommandForAdd
     RemoteCommandTargeter* targeter,
     const std::string& dbName,
     const BSONObj& cmdObj) {
-    auto host = targeter->findHost(ReadPreferenceSetting{ReadPreference::PrimaryOnly},
-                                   RemoteCommandTargeter::selectFindHostMaxWaitTime(txn));
+    auto host = targeter->findHost(txn, ReadPreferenceSetting{ReadPreference::PrimaryOnly});
     if (!host.isOK()) {
         return host.getStatus();
     }
@@ -1712,8 +1711,8 @@ void ShardingCatalogManagerImpl::_scheduleAddShardTask(
 
     // Schedule the shardIdentity upsert request to run immediately, and track the handle.
 
-    auto swHost = targeter->findHost(ReadPreferenceSetting{ReadPreference::PrimaryOnly},
-                                     Milliseconds(kDefaultFindHostMaxWaitTime));
+    auto swHost = targeter->findHostWithMaxWait(ReadPreferenceSetting{ReadPreference::PrimaryOnly},
+                                                Milliseconds(kDefaultFindHostMaxWaitTime));
     if (!swHost.isOK()) {
         // A 3.2 mongos must have previously successfully communicated with hosts in this shard,
         // so a failure to find a host here is probably transient, and it is safe to retry.
diff --git a/src/mongo/s/client/shard_remote.cpp b/src/mongo/s/client/shard_remote.cpp
index 3dbae46be66..200f3fc30ad 100644
--- a/src/mongo/s/client/shard_remote.cpp
+++ b/src/mongo/s/client/shard_remote.cpp
@@ -176,8 +176,7 @@ Shard::HostWithResponse ShardRemote::_runCommand(OperationContext* txn,
     if (getId() == "config") {
         readPrefWithMinOpTime.minOpTime = grid.configOpTime();
     }
-    const auto host = _targeter->findHost(readPrefWithMinOpTime,
-                                          RemoteCommandTargeter::selectFindHostMaxWaitTime(txn));
+    const auto host = _targeter->findHost(txn, readPrefWithMinOpTime);
     if (!host.isOK()) {
         return Shard::HostWithResponse(boost::none, host.getStatus());
     }
@@ -244,8 +243,7 @@ StatusWith<Shard::QueryResponse> ShardRemote::_exhaustiveFindOnConfig(
     ReadPreferenceSetting readPrefWithMinOpTime(readPref);
     readPrefWithMinOpTime.minOpTime = grid.configOpTime();
 
-    const auto host = _targeter->findHost(readPrefWithMinOpTime,
-                                          RemoteCommandTargeter::selectFindHostMaxWaitTime(txn));
+    const auto host = _targeter->findHost(txn, readPrefWithMinOpTime);
     if (!host.isOK()) {
         return host.getStatus();
     }
diff --git a/src/mongo/s/commands/cluster_get_last_error_cmd.cpp b/src/mongo/s/commands/cluster_get_last_error_cmd.cpp
index 56bfa66b369..dbeb35a673e 100644
--- a/src/mongo/s/commands/cluster_get_last_error_cmd.cpp
+++ b/src/mongo/s/commands/cluster_get_last_error_cmd.cpp
@@ -117,7 +117,7 @@ public:
                 break;
             }
             auto shard = shardStatus.getValue();
-            auto swHostAndPort = shard->getTargeter()->findHost(readPref);
+            auto swHostAndPort = shard->getTargeter()->findHostNoWait(readPref);
             if (!swHostAndPort.isOK()) {
                 status = swHostAndPort.getStatus();
                 break;
diff --git a/src/mongo/s/commands/cluster_write_cmd.cpp b/src/mongo/s/commands/cluster_write_cmd.cpp
index ea63a7d020a..925edaddf62 100644
--- a/src/mongo/s/commands/cluster_write_cmd.cpp
+++ b/src/mongo/s/commands/cluster_write_cmd.cpp
@@ -271,7 +271,7 @@ private:
             if (!shardStatus.isOK()) {
                 return shardStatus.getStatus();
             }
-            auto swHostAndPort = shardStatus.getValue()->getTargeter()->findHost(readPref);
+            auto swHostAndPort = shardStatus.getValue()->getTargeter()->findHostNoWait(readPref);
             if (!swHostAndPort.isOK()) {
                 return swHostAndPort.getStatus();
             }
diff --git a/src/mongo/s/query/async_results_merger.cpp b/src/mongo/s/query/async_results_merger.cpp
index 85cc8454dd4..c95a0fe2482 100644
--- a/src/mongo/s/query/async_results_merger.cpp
+++ b/src/mongo/s/query/async_results_merger.cpp
@@ -685,8 +685,7 @@ Status AsyncResultsMerger::RemoteCursorData::resolveShardIdToHostAndPort(
     }
 
     // TODO: Pass down an OperationContext* to use here.
-    auto findHostStatus = shard->getTargeter()->findHost(
-        readPref, RemoteCommandTargeter::selectFindHostMaxWaitTime(nullptr));
+    auto findHostStatus = shard->getTargeter()->findHostWithMaxWait(readPref, Seconds{20});
     if (!findHostStatus.isOK()) {
         return findHostStatus.getStatus();
     }
diff --git a/src/mongo/s/server.cpp b/src/mongo/s/server.cpp
index cd98e0cc6dc..79b149de6d7 100644
--- a/src/mongo/s/server.cpp
+++ b/src/mongo/s/server.cpp
@@ -131,6 +131,7 @@ static ExitCode initService();
 // prior execution of mongo initializers or the existence of threads.
 static void cleanupTask() {
     {
+        auto serviceContext = getGlobalServiceContext();
         Client::initThreadIfNotAlready();
         Client& client = cc();
         ServiceContext::UniqueOperationContext uniqueTxn;
@@ -140,6 +141,9 @@ static void cleanupTask() {
             txn = uniqueTxn.get();
         }
 
+        if (serviceContext)
+            serviceContext->setKillAllOperations();
+
         auto cursorManager = grid.getCursorManager();
         cursorManager->shutdown();
         grid.getExecutorPool()->shutdownAndJoin();
diff --git a/src/mongo/s/write_ops/batch_write_exec.cpp b/src/mongo/s/write_ops/batch_write_exec.cpp
index 34f399267f4..6a6395eea13 100644
--- a/src/mongo/s/write_ops/batch_write_exec.cpp
+++ b/src/mongo/s/write_ops/batch_write_exec.cpp
@@ -190,7 +190,7 @@ void BatchWriteExec::executeBatch(OperationContext* txn,
                 } else {
                     auto shard = shardStatus.getValue();
 
-                    auto swHostAndPort = shard->getTargeter()->findHost(readPref);
+                    auto swHostAndPort = shard->getTargeter()->findHostNoWait(readPref);
                     if (!swHostAndPort.isOK()) {
 
                         // Record a resolve failure