Avoid sleep in DBTest.GroupCommitTest to fix flakiness

Summary:
DBTest.GroupCommitTest would often fail when run under valgrind because its sleeps were insufficient to guarantee a group commit had multiple entries. Instead we can use sync point to force a leader to wait until a non-leader thread has enqueued its work, thus guaranteeing a leader can do group commit work for multiple threads.
Closes https://github.com/facebook/rocksdb/pull/3883

Differential Revision: D8079429

Pulled By: ajkr

fbshipit-source-id: 61dc50fad29d2c85547842f681288de60fa29049

diff --git a/db/db_impl_write.cc b/db/db_impl_write.cc
index 9b43bd5fc4af5b92d8d196a54799681ae8c9c288..fce910c708cc85afb8633fcfcd7e0ed5beb34ac2 100644 (file)
--- a/db/db_impl_write.cc
+++ b/db/db_impl_write.cc
@@ -211,6 +211,7 @@ Status DBImpl::WriteImpl(const WriteOptions& write_options,
   // and protects against concurrent loggers and concurrent writes
   // into memtables
 
+  TEST_SYNC_POINT("DBImpl::WriteImpl:BeforeLeaderEnters");
   last_batch_group_size_ =
       write_thread_.EnterAsBatchGroupLeader(&w, &write_group);
 

diff --git a/db/db_test.cc b/db/db_test.cc
index c2dce2b3adad1c910256ab7ed666569cbbb76e3e..ac17606c161214b49abdb054a9256f9697220f11 100644 (file)
--- a/db/db_test.cc
+++ b/db/db_test.cc
@@ -2153,10 +2153,14 @@ TEST_F(DBTest, GroupCommitTest) {
   do {
     Options options = CurrentOptions();
     options.env = env_;
-    env_->log_write_slowdown_.store(100);
     options.statistics = rocksdb::CreateDBStatistics();
     Reopen(options);
 
+    rocksdb::SyncPoint::GetInstance()->LoadDependency(
+        {{"WriteThread::JoinBatchGroup:BeganWaiting",
+          "DBImpl::WriteImpl:BeforeLeaderEnters"}});
+    rocksdb::SyncPoint::GetInstance()->EnableProcessing();
+
     // Start threads
     GCThread thread[kGCNumThreads];
     for (int id = 0; id < kGCNumThreads; id++) {
@@ -2165,13 +2169,7 @@ TEST_F(DBTest, GroupCommitTest) {
       thread[id].done = false;
       env_->StartThread(GCThreadBody, &thread[id]);
     }
-
-    for (int id = 0; id < kGCNumThreads; id++) {
-      while (thread[id].done == false) {
-        env_->SleepForMicroseconds(100000);
-      }
-    }
-    env_->log_write_slowdown_.store(0);
+    env_->WaitForJoin();
 
     ASSERT_GT(TestGetTickerCount(options, WRITE_DONE_BY_OTHER), 0);
 

diff --git a/db/write_thread.cc b/db/write_thread.cc
index f35e0dd21ad8875226eee071dadb634288c0d0f3..c58dbac0db8cb73dfc9f9f945c3201c35015b1ba 100644 (file)
--- a/db/write_thread.cc
+++ b/db/write_thread.cc
@@ -318,6 +318,7 @@ void WriteThread::JoinBatchGroup(Writer* w) {
      * 3.2) an existing memtable writer group leader tell us to finish memtable
      *      writes in parallel.
      */
+    TEST_SYNC_POINT_CALLBACK("WriteThread::JoinBatchGroup:BeganWaiting", w);
     AwaitState(w, STATE_GROUP_LEADER | STATE_MEMTABLE_WRITER_LEADER |
                       STATE_PARALLEL_MEMTABLE_WRITER | STATE_COMPLETED,
                &jbg_ctx);