add EMode to WorkerTheadPool to avoid thread starvation (#492)

- Improvement: Add a new mode to worker thread pools to avoid starvation of workers which could cause long stalls due to other work begin queued up. UE-305498

1 files changed, 59 insertions, 45 deletions

diff --git a/src/zenstore/cache/structuredcachestore.cpp b/src/zenstore/cache/structuredcachestore.cpp
index 1f2d6c37f..3f27e6d21 100644
--- a/src/zenstore/cache/structuredcachestore.cpp
+++ b/src/zenstore/cache/structuredcachestore.cpp
@@ -1574,10 +1574,12 @@ TEST_CASE("cachestore.threadedinsert")	// * doctest::skip(true))
 			std::atomic<size_t> WorkCompleted = 0;
 			for (const auto& Chunk : Chunks)
 			{
-				ThreadPool.ScheduleWork([&Zcs, &WorkCompleted, &Chunk]() {
-					Zcs.Put(Chunk.second.Bucket, Chunk.first, {.Value = Chunk.second.Buffer}, {}, false);
-					WorkCompleted.fetch_add(1);
-				});
+				ThreadPool.ScheduleWork(
+					[&Zcs, &WorkCompleted, &Chunk]() {
+						Zcs.Put(Chunk.second.Bucket, Chunk.first, {.Value = Chunk.second.Buffer}, {}, false);
+						WorkCompleted.fetch_add(1);
+					},
+					WorkerThreadPool::EMode::EnableBacklog);
 			}
 			while (WorkCompleted < Chunks.size())
 			{
@@ -1612,16 +1614,18 @@ TEST_CASE("cachestore.threadedinsert")	// * doctest::skip(true))
 			std::atomic<size_t> WorkCompleted = 0;
 			for (const auto& Chunk : Chunks)
 			{
-				ThreadPool.ScheduleWork([&Zcs, &WorkCompleted, &Chunk]() {
-					std::string	  Bucket	= Chunk.second.Bucket;
-					IoHash		  ChunkHash = Chunk.first;
-					ZenCacheValue CacheValue;
+				ThreadPool.ScheduleWork(
+					[&Zcs, &WorkCompleted, &Chunk]() {
+						std::string	  Bucket	= Chunk.second.Bucket;
+						IoHash		  ChunkHash = Chunk.first;
+						ZenCacheValue CacheValue;
 
-					CHECK(Zcs.Get(Bucket, ChunkHash, CacheValue));
-					IoHash Hash = IoHash::HashBuffer(CacheValue.Value);
-					CHECK(ChunkHash == Hash);
-					WorkCompleted.fetch_add(1);
-				});
+						CHECK(Zcs.Get(Bucket, ChunkHash, CacheValue));
+						IoHash Hash = IoHash::HashBuffer(CacheValue.Value);
+						CHECK(ChunkHash == Hash);
+						WorkCompleted.fetch_add(1);
+					},
+					WorkerThreadPool::EMode::EnableBacklog);
 			}
 			while (WorkCompleted < Chunks.size())
 			{
@@ -1655,23 +1659,27 @@ TEST_CASE("cachestore.threadedinsert")	// * doctest::skip(true))
 			std::atomic_uint32_t AddedChunkCount = 0;
 			for (const auto& Chunk : NewChunks)
 			{
-				ThreadPool.ScheduleWork([&Zcs, &WorkCompleted, Chunk, &AddedChunkCount]() {
-					Zcs.Put(Chunk.second.Bucket, Chunk.first, {.Value = Chunk.second.Buffer}, {}, false);
-					AddedChunkCount.fetch_add(1);
-					WorkCompleted.fetch_add(1);
-				});
+				ThreadPool.ScheduleWork(
+					[&Zcs, &WorkCompleted, Chunk, &AddedChunkCount]() {
+						Zcs.Put(Chunk.second.Bucket, Chunk.first, {.Value = Chunk.second.Buffer}, {}, false);
+						AddedChunkCount.fetch_add(1);
+						WorkCompleted.fetch_add(1);
+					},
+					WorkerThreadPool::EMode::EnableBacklog);
 			}
 
 			for (const auto& Chunk : Chunks)
 			{
-				ThreadPool.ScheduleWork([&Zcs, &WorkCompleted, Chunk]() {
-					ZenCacheValue CacheValue;
-					if (Zcs.Get(Chunk.second.Bucket, Chunk.first, CacheValue))
-					{
-						CHECK(Chunk.first == IoHash::HashBuffer(CacheValue.Value));
-					}
-					WorkCompleted.fetch_add(1);
-				});
+				ThreadPool.ScheduleWork(
+					[&Zcs, &WorkCompleted, Chunk]() {
+						ZenCacheValue CacheValue;
+						if (Zcs.Get(Chunk.second.Bucket, Chunk.first, CacheValue))
+						{
+							CHECK(Chunk.first == IoHash::HashBuffer(CacheValue.Value));
+						}
+						WorkCompleted.fetch_add(1);
+					},
+					WorkerThreadPool::EMode::EnableBacklog);
 			}
 			while (AddedChunkCount.load() < NewChunks.size())
 			{
@@ -1710,12 +1718,14 @@ TEST_CASE("cachestore.threadedinsert")	// * doctest::skip(true))
 				std::atomic<size_t> WorkCompleted = 0;
 				for (const auto& Chunk : GcChunkHashes)
 				{
-					ThreadPool.ScheduleWork([&Zcs, &WorkCompleted, Chunk]() {
-						ZenCacheValue CacheValue;
-						CHECK(Zcs.Get(Chunk.second, Chunk.first, CacheValue));
-						CHECK(Chunk.first == IoHash::HashBuffer(CacheValue.Value));
-						WorkCompleted.fetch_add(1);
-					});
+					ThreadPool.ScheduleWork(
+						[&Zcs, &WorkCompleted, Chunk]() {
+							ZenCacheValue CacheValue;
+							CHECK(Zcs.Get(Chunk.second, Chunk.first, CacheValue));
+							CHECK(Chunk.first == IoHash::HashBuffer(CacheValue.Value));
+							WorkCompleted.fetch_add(1);
+						},
+						WorkerThreadPool::EMode::EnableBacklog);
 				}
 				while (WorkCompleted < GcChunkHashes.size())
 				{
@@ -1848,11 +1858,13 @@ TEST_CASE("cachestore.drop.bucket")
 		CHECK(Value1.Value);
 
 		std::atomic_bool WorkComplete = false;
-		Workers.ScheduleWork([&]() {
-			zen::Sleep(100);
-			Value1.Value = IoBuffer{};
-			WorkComplete = true;
-		});
+		Workers.ScheduleWork(
+			[&]() {
+				zen::Sleep(100);
+				Value1.Value = IoBuffer{};
+				WorkComplete = true;
+			},
+			WorkerThreadPool::EMode::EnableBacklog);
 		// On Windows, DropBucket() will be blocked as long as we hold a reference to a buffer in the bucket
 		// Our DropBucket execution blocks any incoming request from completing until we are done with the drop
 		CHECK(Zcs.DropBucket(Namespace, Bucket));
@@ -1931,14 +1943,16 @@ TEST_CASE("cachestore.drop.namespace")
 		CHECK(Value4.Value);
 
 		std::atomic_bool WorkComplete = false;
-		Workers.ScheduleWork([&]() {
-			zen::Sleep(100);
-			Value1.Value = IoBuffer{};
-			Value2.Value = IoBuffer{};
-			Value3.Value = IoBuffer{};
-			Value4.Value = IoBuffer{};
-			WorkComplete = true;
-		});
+		Workers.ScheduleWork(
+			[&]() {
+				zen::Sleep(100);
+				Value1.Value = IoBuffer{};
+				Value2.Value = IoBuffer{};
+				Value3.Value = IoBuffer{};
+				Value4.Value = IoBuffer{};
+				WorkComplete = true;
+			},
+			WorkerThreadPool::EMode::EnableBacklog);
 		// On Windows, DropBucket() will be blocked as long as we hold a reference to a buffer in the bucket
 		// Our DropBucket execution blocks any incoming request from completing until we are done with the drop
 		CHECK(Zcs.DropNamespace(Namespace1));