sk0x50 commented on a change in pull request #5779: IGNITE-10508 Support the
new checkpoint feature not wait for the previous operation to complete
URL: https://github.com/apache/ignite/pull/5779#discussion_r253447989
##########
File path:
modules/core/src/test/java/org/apache/ignite/internal/processors/cache/persistence/IgnitePdsCheckpointCancelTest.java
##########
@@ -0,0 +1,579 @@
+package org.apache.ignite.internal.processors.cache.persistence;
+
+import java.io.File;
+import java.io.IOException;
+import java.nio.ByteBuffer;
+import java.nio.file.OpenOption;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Map;
+import java.util.concurrent.ConcurrentHashMap;
+import java.util.concurrent.CountDownLatch;
+import java.util.concurrent.atomic.AtomicBoolean;
+import java.util.concurrent.atomic.AtomicLong;
+import java.util.function.Function;
+import org.apache.ignite.Ignite;
+import org.apache.ignite.IgniteDataStreamer;
+import org.apache.ignite.IgniteException;
+import org.apache.ignite.cache.affinity.rendezvous.RendezvousAffinityFunction;
+import org.apache.ignite.configuration.CacheConfiguration;
+import org.apache.ignite.configuration.DataRegionConfiguration;
+import org.apache.ignite.configuration.DataStorageConfiguration;
+import org.apache.ignite.configuration.IgniteConfiguration;
+import org.apache.ignite.internal.IgniteEx;
+import org.apache.ignite.internal.IgniteInterruptedCheckedException;
+import
org.apache.ignite.internal.processors.cache.persistence.GridCacheDatabaseSharedManager.CheckpointProgress;
+import
org.apache.ignite.internal.processors.cache.persistence.file.AsyncFileIOFactory;
+import org.apache.ignite.internal.processors.cache.persistence.file.FileIO;
+import
org.apache.ignite.internal.processors.cache.persistence.file.FileIODecorator;
+import
org.apache.ignite.internal.processors.cache.persistence.file.FileIOFactory;
+import
org.apache.ignite.internal.processors.cache.persistence.pagemem.PageMemoryImpl;
+import org.apache.ignite.internal.util.typedef.internal.CU;
+import org.apache.ignite.internal.util.typedef.internal.U;
+import org.apache.ignite.testframework.junits.common.GridCommonAbstractTest;
+import org.junit.After;
+import org.junit.Assert;
+import org.junit.Before;
+import org.junit.Test;
+import org.junit.runner.RunWith;
+import org.junit.runners.JUnit4;
+
+import static
org.apache.ignite.internal.processors.cache.persistence.file.FilePageStoreManager.PART_FILE_PREFIX;
+
+@RunWith(JUnit4.class)
+public class IgnitePdsCheckpointCancelTest extends GridCommonAbstractTest {
+ private static final String NODE_CONST_ID = "NODE";
+ private static final String REGION_NAME = "MY_REGION";
+
+ private static final int PAGE_SIZE = 4096;
+ private static final int REGION_SIZE = 400 * 1024 * 1024;
+ private static final int CP_BUFFER_SIZE = 400 * 1024 * 1024;
+
+ private static final int DIRTY_PAGES_BOUND = REGION_SIZE / 2;
+
+ private static final int MIDDLE_PAGE_COUNT = DIRTY_PAGES_BOUND / PAGE_SIZE
/ 4;
+
+ private List<TestFileIO> partsFilesIO = new ArrayList<>();
+
+ private FileIOFactory ioFactory;
+
+ private boolean cancelCp = true;
+
+ private int partitions;
+
+ private final TimeTracker time = new TimeTracker();
+
+ @Override
+ protected IgniteConfiguration getConfiguration(String name) throws
Exception {
+ assert partitions != 0;
+
+ IgniteConfiguration cfg = super.getConfiguration(name);
+
+ cfg.setConsistentId(NODE_CONST_ID);
+
+ cfg.setDataStorageConfiguration(
+ new DataStorageConfiguration()
+ .setFileIOFactory(ioFactory != null ? ioFactory : new
AsyncFileIOFactory())
+ .setCheckpointFrequency(60 * 60 * 1000)
+ .setWalSegments(10)
+ .setWalSegmentSize(REGION_SIZE)
+ .setMaxWalArchiveSize(REGION_SIZE * 10L)
+ .setDefaultDataRegionConfiguration(
+ new DataRegionConfiguration()
+ .setName(REGION_NAME)
+ .setMaxSize(REGION_SIZE)
+ .setCheckpointPageBufferSize(CP_BUFFER_SIZE)
+ .setMetricsEnabled(true)
+ .setPersistenceEnabled(true)
+ )
+ );
+
+ cfg.setCacheConfiguration(
+ new CacheConfiguration(DEFAULT_CACHE_NAME)
+ .setDataRegionName(REGION_NAME)
+ .setAffinity(new RendezvousAffinityFunction(false, partitions))
+ );
+
+ return cfg;
+ }
+
+ @Before
+ public void before() throws Exception {
+ stopAllGrids();
+
+ cleanPersistenceDir();
+
+ }
+
+ @After
+ public void after() throws Exception {
+ stopAllGrids();
+
+ cleanPersistenceDir();
+ }
+
+ public void testCancelOnMarkCheckpointBegin() {
+ //TODO
+ }
+
+ @Test
+ public void testCancelOnPageWrite() throws Exception {
+ partitions = 1;
+ ioFactory = new TestFileIOFactory(new AsyncFileIOFactory(),
partsFilesIO);
+
+ String firstCpTime = "firstCpTime";
+ String awaitMiddleWritePagesInCheckpoint =
"awaitMiddleWritePagesInCheckpoint";
+ String secondCpAwaitBegin = "secondCpAwaitBeginTime";
+
+ Ignite ig = startGrid(0);
+
+ ig.cluster().active(true);
+
+ int key = 0;
+
+ try (IgniteDataStreamer<Integer, byte[]> st =
ig.dataStreamer(DEFAULT_CACHE_NAME)) {
+ st.allowOverwrite(true);
+ st.perNodeBufferSize(10);
+ st.perThreadBufferSize(10);
+
+ long allocatedPages;
+
+ for (; ; ) {
+ allocatedPages =
ig.dataRegionMetrics(REGION_NAME).getTotalAllocatedPages();
+
+ if (allocatedPages >= (MIDDLE_PAGE_COUNT * 3))
+ break;
+
+ st.addData(key++, new byte[PAGE_SIZE]);
+ }
+
+ log.info("Added " + key + " key value pairs, allocatedPages=" +
+ allocatedPages + ", bound=" + (MIDDLE_PAGE_COUNT * 3));
+ }
+
+ log.info("\nConfig: \n" +
+ "PAGES_IN_NEXT_CP:" +
ig.dataRegionMetrics(REGION_NAME).getTotalAllocatedPages() + "\n" +
+ "REGION_SIZE:" + REGION_SIZE + "(" + (REGION_SIZE / PAGE_SIZE) +
")\n" +
+ "DIRTY_PAGES_BOUND:" + DIRTY_PAGES_BOUND + "(" +
(DIRTY_PAGES_BOUND / PAGE_SIZE) + ")\n" +
+ "MIDDLE_PAGE_COUNT:" + MIDDLE_PAGE_COUNT + "\n"
+ );
+
+ GridCacheDatabaseSharedManager dbMgr = dbManager(ig);
+
+ TestFileIO.reset();
+
+ CheckpointFuture cp1Future = dbMgr.forceCheckpoint("chp=1", cancelCp);
+
+ cp1Future.beginFuture().get();
+
+ time.begin(firstCpTime);
+
+ cp1Future.finishFuture()
+ .listen(f -> {
+ time.end(firstCpTime);
+
+ log.info("chp=1 time duration (" + time.getByName(firstCpTime)
+ ")");
+ });
+
+ time.begin(awaitMiddleWritePagesInCheckpoint);
+
+ TestFileIO.await(MIDDLE_PAGE_COUNT);
+
+ time.end(awaitMiddleWritePagesInCheckpoint);
+
+ log.info("Done to wait middle page write in current checkpoint,
pages:" +
+ TestFileIO.SHARED_PAGE_COUNTER.get());
+
+ Assert.assertTrue(!cp1Future.finishFuture().isDone());
+
+ time.begin(secondCpAwaitBegin);
+
+ dbMgr.forceCheckpoint("chp=2, chp=1 canceled",
cancelCp).beginFuture().get();
+
+ time.end(secondCpAwaitBegin);
+
+ long secondCpAwaitBeginTime = time.getByName(secondCpAwaitBegin);
+ long awaitMiddleWritePagesInCheckpointTime =
time.getByName(awaitMiddleWritePagesInCheckpoint);
+
+ String msg = "waitNextCpTime=" + secondCpAwaitBeginTime +
+ ", writePagesInTheMiddleTime=" +
awaitMiddleWritePagesInCheckpointTime;
+
+ Assert.assertTrue(msg, secondCpAwaitBeginTime <
awaitMiddleWritePagesInCheckpointTime);
+
+ log.info(msg);
+ }
+
+ @Test
+ public void testCancelOnSyncStores() throws Exception {
+ partitions = 1024;
+
+ String firstCpTimeLable = "firstCpTime";
+ String secondCpAwaitBeginLabel = "secondCpAwaitBeginTime";
+
+ Ignite ig = startGrid(0);
+
+ ig.cluster().active(true);
+
+ int key = 0;
+
+ try (IgniteDataStreamer<Integer, byte[]> st =
ig.dataStreamer(DEFAULT_CACHE_NAME)) {
+ st.allowOverwrite(true);
+ st.perNodeBufferSize(10);
+ st.perThreadBufferSize(10);
+
+ long allocatedPages;
+
+ for (; ; ) {
+ allocatedPages =
ig.dataRegionMetrics(REGION_NAME).getTotalAllocatedPages();
+
+ if (allocatedPages >= (MIDDLE_PAGE_COUNT * 3))
+ break;
+
+ st.addData(key++, new byte[PAGE_SIZE]);
+ }
+
+ log.info("Added " + key + " key value pairs, allocatedPages=" +
+ allocatedPages + ", bound=" + (MIDDLE_PAGE_COUNT * 3));
+ }
+
+ log.info("\nConfig: \n" +
+ "PAGES_IN_NEXT_CP:" +
ig.dataRegionMetrics(REGION_NAME).getTotalAllocatedPages() + "\n" +
+ "REGION_SIZE:" + REGION_SIZE + "(" + (REGION_SIZE / PAGE_SIZE) +
")\n" +
+ "DIRTY_PAGES_BOUND:" + DIRTY_PAGES_BOUND + "(" +
(DIRTY_PAGES_BOUND / PAGE_SIZE) + ")\n" +
+ "MIDDLE_PAGE_COUNT:" + MIDDLE_PAGE_COUNT + "\n"
+ );
+
+ GridCacheDatabaseSharedManager dbMgr = dbManager(ig);
+
+ GridCacheDatabaseSharedManager.Checkpointer checkpointer =
U.field(dbMgr, "checkpointer");
+
+ Function<GridCacheDatabaseSharedManager.Checkpoint, Boolean> prevFunc
= checkpointer.checkCancel;
+
+ CountDownLatch awaitSyncStores = new CountDownLatch(1);
+ CountDownLatch awaitContinueSyncStores = new CountDownLatch(1);
+
+ AtomicBoolean readyToSync = new AtomicBoolean();
+
+ AtomicLong lastCheckTime = new AtomicLong();
+ AtomicLong maxCheckCancelTime = new AtomicLong();
+
+ checkpointer.checkCancel = (chp) -> {
+ CheckpointProgress cpProgress = chp.checkpointProgress();
+
+ long last = lastCheckTime.getAndSet(System.currentTimeMillis());
+
+ if (last != 0) {
+ long lastCheck = System.currentTimeMillis() - last;
+
+ if (lastCheck > maxCheckCancelTime.get())
+ maxCheckCancelTime.set(lastCheck);
+ }
+
+ int stores = cpProgress.pageStores().size();
+
+ // All pages written to store and collection size is equals to
partitions.
+ if (stores == partitions)
+ readyToSync.set(true);
+
+ try {
+ U.sleep(5);
+ }
+ catch (IgniteInterruptedCheckedException e) {
+ Thread.currentThread().interrupt();
+
+ throw new IgniteException(e);
+ }
+
+ if (readyToSync.get()) {
+ if (stores < partitions / 10) {
+ readyToSync.set(false);
+
+ awaitSyncStores.countDown();
+
+ try {
+ log.info("Done to wait store syncs in current
checkpoint, stores:" + stores);
+
+ awaitContinueSyncStores.await();
+ }
+ catch (InterruptedException e) {
+ Thread.currentThread().interrupt();
+
+ throw new IgniteException(e);
+ }
+ }
+ }
+
+ return prevFunc.apply(chp);
+ };
+
+ CheckpointFuture cp1Future = dbMgr.forceCheckpoint("chp=1", cancelCp);
+
+ cp1Future.beginFuture().get();
+
+ time.begin(firstCpTimeLable);
+
+ cp1Future.finishFuture()
+ .listen(f -> {
+ time.end(firstCpTimeLable);
+
+ log.info("chp=1 time duration (" +
time.getByName(firstCpTimeLable) + ")");
+ });
+
+ awaitSyncStores.await();
+
+ Assert.assertTrue(!cp1Future.finishFuture().isDone());
+
+ time.begin(secondCpAwaitBeginLabel);
+
+ CheckpointFuture chpFuture = dbMgr.forceCheckpoint("chp=2, chp=1
canceled", cancelCp);
+
+ awaitContinueSyncStores.countDown();
+
+ chpFuture.beginFuture().get();
+
+ time.end(secondCpAwaitBeginLabel);
+
+ chpFuture.finishFuture().get();
+
+ long secondCpAwaitBeginTime = time.getByName(secondCpAwaitBeginLabel);
+
+ String msg = "waitNextCpTime=" + secondCpAwaitBeginTime + ",
maxCheckCancelTime=" + maxCheckCancelTime.get();
+
+ Assert.assertTrue(msg, secondCpAwaitBeginTime <
maxCheckCancelTime.get() * 2);
+
+ log.info(msg);
+ }
+
+ @Test
+ public void testCancelOnDestroyPartitions() throws Exception {
+ partitions = 1024;
+
+ String firstCpTimeLable = "firstCpTime";
+ String secondCpAwaitBeginLabel = "secondCpAwaitBeginTime";
+
+ Ignite ig = startGrid(0);
+
+ ig.cluster().active(true);
+
+ GridCacheDatabaseSharedManager dbMgr = dbManager(ig);
+
+ GridCacheDatabaseSharedManager.Checkpointer checkpointer =
U.field(dbMgr, "checkpointer");
+
+ Function<GridCacheDatabaseSharedManager.Checkpoint, Boolean> prevFunc
= checkpointer.checkCancel;
+
+ CountDownLatch awaitDestroyParts = new CountDownLatch(1);
+ CountDownLatch awaitContinueDestroyParts = new CountDownLatch(1);
+
+ AtomicLong lastCheckTime = new AtomicLong();
+ AtomicLong maxCheckCancelTime = new AtomicLong();
+
+ checkpointer.checkCancel = (chp) -> {
+ CheckpointProgress cpProgress = chp.checkpointProgress();
+
+ long last = lastCheckTime.getAndSet(System.currentTimeMillis());
+
+ if (last != 0) {
+ long lastCheck = System.currentTimeMillis() - last;
+
+ if (lastCheck > maxCheckCancelTime.get())
+ maxCheckCancelTime.set(lastCheck);
+ }
+
+ int destroyQueueSize = cpProgress.destroyQueue().size();
+
+ try {
+ U.sleep(5);
+ }
+ catch (IgniteInterruptedCheckedException e) {
+ Thread.currentThread().interrupt();
+
+ throw new IgniteException(e);
+ }
+
+ if ((destroyQueueSize < partitions - (partitions / 10)) &&
awaitDestroyParts.getCount() > 0) {
+ awaitDestroyParts.countDown();
+
+ try {
+ log.info("Done to wait destory partitions in current
checkpoint, parts:" + destroyQueueSize);
+
+ awaitContinueDestroyParts.await();
+ }
+ catch (InterruptedException e) {
+ Thread.currentThread().interrupt();
+
+ throw new IgniteException(e);
+ }
+ }
+
+ return prevFunc.apply(chp);
+ };
+
+ for (int i = 0; i < partitions; i++) {
+ int grpId = CU.cacheId(DEFAULT_CACHE_NAME);
+
+ pageMemory(ig, DEFAULT_CACHE_NAME).invalidate(grpId, i);
+
+ dbMgr.schedulePartitionDestroy(grpId, i);
+ }
+
+ CheckpointFuture cp1Future = dbMgr.forceCheckpoint("chp=1", cancelCp);
+
+ cp1Future.beginFuture().get();
+
+ time.begin(firstCpTimeLable);
+
+ cp1Future.finishFuture()
+ .listen(f -> {
+ time.end(firstCpTimeLable);
+
+ log.info("chp=1 time duration (" +
time.getByName(firstCpTimeLable) + ")");
+ });
+
+ awaitDestroyParts.await();
+
+ Assert.assertTrue(!cp1Future.finishFuture().isDone());
+
+ time.begin(secondCpAwaitBeginLabel);
+
+ CheckpointFuture chpFuture = dbMgr.forceCheckpoint("chp=2, chp=1
canceled", cancelCp);
+
+ awaitContinueDestroyParts.countDown();
+
+ chpFuture.beginFuture().get();
+
+ time.end(secondCpAwaitBeginLabel);
+
+ chpFuture.finishFuture().get();
+
+ long secondCpAwaitBeginTime = time.getByName(secondCpAwaitBeginLabel);
+
+ String msg = "waitNextCpTime=" + secondCpAwaitBeginTime + ",
maxCheckCancelTime=" + maxCheckCancelTime.get();
+
+ Assert.assertTrue(msg, secondCpAwaitBeginTime <
maxCheckCancelTime.get() * 4);
+
+ log.info(msg);
+ }
+
+ @Test
+ public void testCancelOnPageWriteAndCrash() {
+
Review comment:
Do we have a ticket to address these tests?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
