Github user eyalfa commented on a diff in the pull request: https://github.com/apache/spark/pull/18855#discussion_r131913940 --- Diff: core/src/test/scala/org/apache/spark/storage/BlockManagerSuite.scala --- @@ -1415,6 +1415,79 @@ class BlockManagerSuite extends SparkFunSuite with Matchers with BeforeAndAfterE super.fetchBlockSync(host, port, execId, blockId) } } + + def testGetOrElseUpdateForLargeBlock(storageLevel : StorageLevel) { + store = makeBlockManager(6L * 1024 * 1024 * 1024, "exec1") + def mkBlobs() = { + val rng = new java.util.Random(42) + val buff = new Array[Byte](1024 * 1024) + rng.nextBytes(buff) + Iterator.fill(2 * 1024 + 1) { + buff + } + } + val res1 = store.getOrElseUpdate( + RDDBlockId(42, 0), + storageLevel, + implicitly[ClassTag[Array[Byte]]], + mkBlobs _ + ) + withClue(res1) { + assert(res1.isLeft) + assert(res1.left.get.data.zipAll(mkBlobs(), null, null).forall { + case (a, b) => + a != null && + b != null && + a.asInstanceOf[Array[Byte]].seq == b.asInstanceOf[Array[Byte]].seq + }) + } + val getResult = store.get(RDDBlockId(42, 0)) + withClue(getResult) { + assert(getResult.isDefined) + assert(getResult.get.data.zipAll(mkBlobs(), null, null).forall { + case (a, b) => + a != null && + b != null && + a.asInstanceOf[Array[Byte]].seq == b.asInstanceOf[Array[Byte]].seq + }) + } + val getBlockRes = store.getBlockData(RDDBlockId(42, 0)) + withClue(getBlockRes) { + try { + assert(getBlockRes.size() >= 2 * 1024 * 1024 * 1024) + Utils.tryWithResource(getBlockRes.createInputStream()) { inpStrm => + val iter = store + .serializerManager + .dataDeserializeStream(RDDBlockId(42, 0) + , inpStrm)(implicitly[ClassTag[Array[Byte]]]) + assert(iter.zipAll(mkBlobs(), null, null).forall { + case (a, b) => + a != null && + b != null && + a.asInstanceOf[Array[Byte]].seq == b.asInstanceOf[Array[Byte]].seq + }) + } + } finally { + getBlockRes.release() + } + } + } + + test("getOrElseUpdate > 2gb, storage level = disk only") { --- End diff -- these tests cover more than just the `DiskOnly` storage level, they were crafted when I had bigger ambitions of solving the entire 2GB issue :sunglasses: , that was before seeing some ~100 files pull requests being abandoned or rejected. aside, these tests also test the entire orchestration done by `BlockManager` when an `RDD` requests a cached partition, notice that these tests intentionally makes two calls to the BlockManager in order to simulate both code paths (cache-hit, cache-miss).
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org