anton-vinogradov commented on code in PR #10199: URL: https://github.com/apache/ignite/pull/10199#discussion_r947873656
########## modules/control-utility/src/test/java/org/apache/ignite/util/GridCommandHandlerConsistencyCountersTest.java: ########## @@ -0,0 +1,575 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package org.apache.ignite.util; + +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.Set; +import java.util.concurrent.ConcurrentHashMap; +import java.util.concurrent.CountDownLatch; +import java.util.concurrent.ThreadLocalRandom; +import java.util.concurrent.atomic.AtomicBoolean; +import java.util.concurrent.atomic.AtomicReference; +import java.util.function.Consumer; +import java.util.regex.Matcher; +import java.util.regex.Pattern; +import org.apache.ignite.Ignite; +import org.apache.ignite.IgniteCache; +import org.apache.ignite.cache.CacheAtomicityMode; +import org.apache.ignite.cache.CachePeekMode; +import org.apache.ignite.cache.ReadRepairStrategy; +import org.apache.ignite.cache.affinity.rendezvous.RendezvousAffinityFunction; +import org.apache.ignite.cluster.ClusterNode; +import org.apache.ignite.cluster.ClusterState; +import org.apache.ignite.configuration.CacheConfiguration; +import org.apache.ignite.configuration.IgniteConfiguration; +import org.apache.ignite.configuration.WALMode; +import org.apache.ignite.failure.StopNodeFailureHandler; +import org.apache.ignite.internal.IgniteEx; +import org.apache.ignite.internal.TestRecordingCommunicationSpi; +import org.apache.ignite.internal.commandline.consistency.ConsistencyCommand; +import org.apache.ignite.internal.processors.cache.distributed.dht.GridDhtTxFinishRequest; +import org.apache.ignite.internal.processors.cache.distributed.dht.GridDhtTxPrepareRequest; +import org.apache.ignite.internal.processors.cache.distributed.dht.atomic.GridDhtAtomicSingleUpdateRequest; +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.util.typedef.F; +import org.apache.ignite.internal.util.typedef.G; +import org.apache.ignite.lang.IgniteBiPredicate; +import org.apache.ignite.plugin.extensions.communication.Message; +import org.apache.ignite.testframework.ListeningTestLogger; +import org.apache.ignite.testframework.LogListener; +import org.junit.Test; +import org.junit.runner.RunWith; +import org.junit.runners.Parameterized; + +import static org.apache.ignite.cache.CacheAtomicityMode.ATOMIC; +import static org.apache.ignite.cache.CacheAtomicityMode.TRANSACTIONAL; +import static org.apache.ignite.cache.CacheWriteSynchronizationMode.FULL_SYNC; +import static org.apache.ignite.cache.ReadRepairStrategy.PRIMARY; +import static org.apache.ignite.cache.ReadRepairStrategy.RELATIVE_MAJORITY; +import static org.apache.ignite.cache.ReadRepairStrategy.REMOVE; +import static org.apache.ignite.internal.commandline.CommandHandler.EXIT_CODE_OK; +import static org.apache.ignite.internal.processors.cache.persistence.GridCacheOffheapManager.DFLT_WAL_MARGIN_FOR_ATOMIC_CACHE_HISTORICAL_REBALANCE; +import static org.apache.ignite.internal.visor.consistency.VisorConsistencyRepairTask.CONSISTENCY_VIOLATIONS_FOUND; +import static org.apache.ignite.internal.visor.consistency.VisorConsistencyRepairTask.NOTHING_FOUND; +import static org.apache.ignite.testframework.GridTestUtils.assertContains; +import static org.apache.ignite.testframework.GridTestUtils.assertNotContains; +import static org.apache.ignite.testframework.LogListener.matches; + +/** + * + */ +@RunWith(Parameterized.class) +public class GridCommandHandlerConsistencyCountersTest extends GridCommandHandlerClusterPerMethodAbstractTest { + /** */ + @Parameterized.Parameters(name = "strategy={0}, reuse={1}, historical={2}, atomicity={3}") + public static Iterable<Object[]> data() { + List<Object[]> res = new ArrayList<>(); + + for (ReadRepairStrategy strategy : ReadRepairStrategy.values()) { + for (boolean reuse : new boolean[] {false, true}) { + for (boolean historical : new boolean[] {false, true}) { + for (CacheAtomicityMode atomicityMode : new CacheAtomicityMode[] {ATOMIC, TRANSACTIONAL}) + res.add(new Object[] {strategy, reuse, historical, atomicityMode}); + } + } + } + + return res; + } + + /** + * ReadRepair strategy + */ + @Parameterized.Parameter + public ReadRepairStrategy strategy; + + /** + * When true, updates will reuse already existing keys. + */ + @Parameterized.Parameter(1) + public boolean reuseKeys; + + /** + * When true, historical rebalance will be used instead of full. + */ + @Parameterized.Parameter(2) + public boolean historical; + + /** + * Cache atomicity mode + */ + @Parameterized.Parameter(3) + public CacheAtomicityMode atomicityMode; + + /** Listening logger. */ + protected final ListeningTestLogger listeningLog = new ListeningTestLogger(log); + + /** File IO blocked flag. */ + private static volatile boolean ioBlocked; + + /** {@inheritDoc} */ + @Override protected void beforeTestsStarted() throws Exception { + super.beforeTestsStarted(); + + cleanPersistenceDir(); + } + + /** {@inheritDoc} */ + @Override protected IgniteConfiguration getConfiguration(String igniteInstanceName) throws Exception { + IgniteConfiguration cfg = super.getConfiguration(igniteInstanceName); + + cfg.setGridLogger(listeningLog); + + cfg.getDataStorageConfiguration().setFileIOFactory( + new BlockableFileIOFactory(cfg.getDataStorageConfiguration().getFileIOFactory())); + + cfg.getDataStorageConfiguration().setWalMode(WALMode.FSYNC); // Allows to use special IO at WAL as well. + + cfg.setFailureHandler(new StopNodeFailureHandler()); // Helps to kill nodes on stop with disabled IO. + + return cfg; + } + + /** + * + */ + private static class BlockableFileIOFactory implements FileIOFactory { + /** IO Factory. */ + private final FileIOFactory factory; + + /** + * @param factory Factory. + */ + public BlockableFileIOFactory(FileIOFactory factory) { + this.factory = factory; + } + + /** {@inheritDoc} */ + @Override public FileIO create(File file, OpenOption... modes) throws IOException { + return new FileIODecorator(factory.create(file, modes)) { + @Override public int write(ByteBuffer srcBuf) throws IOException { + if (ioBlocked) + throw new IOException(); + + return super.write(srcBuf); + } + + @Override public int write(ByteBuffer srcBuf, long position) throws IOException { + if (ioBlocked) + throw new IOException(); + + return super.write(srcBuf, position); + } + + @Override public int write(byte[] buf, int off, int len) throws IOException { + if (ioBlocked) + throw new IOException(); + + return super.write(buf, off, len); + } + }; + } + } + + /** + * Checks counters and behaviour on crash recovery. + */ + @Test + public void testCountersOnCrachRecovery() throws Exception { + int nodes = 3; + int backupNodes = nodes - 1; + + IgniteEx ignite = startGrids(nodes); + + ignite.cluster().state(ClusterState.ACTIVE); + + IgniteCache<Object, Object> cache = ignite.createCache(new CacheConfiguration<>() + .setAffinity(new RendezvousAffinityFunction(false, 1)) + .setBackups(backupNodes) + .setName(DEFAULT_CACHE_NAME) + .setAtomicityMode(atomicityMode) + .setWriteSynchronizationMode(FULL_SYNC) // Allows to be sure that all messages are sent when put succeed. + .setReadFromBackup(true)); // Allows to check values on backups. + + int updateCnt = 0; + + // Initial preloading. + for (int i = 0; i < 2_000; i++) { // Enough to have historical rebalance when needed. + cache.put(i, i); + + updateCnt++; + } + + // Trick to have historical rebalance on cluster recovery (decreases percent of updates in comparison to cache size). + if (historical) { + stopAllGrids(); + startGrids(nodes); + } + + int preloadCnt = updateCnt; + + Ignite prim = primaryNode(0L, DEFAULT_CACHE_NAME); + List<Ignite> backups = backupNodes(0L, DEFAULT_CACHE_NAME); + + AtomicBoolean prepareBlock = new AtomicBoolean(); + AtomicBoolean finishBlock = new AtomicBoolean(); + + AtomicReference<CountDownLatch> blockLatch = new AtomicReference<>(); + + TestRecordingCommunicationSpi.spi(prim).blockMessages(new IgniteBiPredicate<ClusterNode, Message>() { + @Override public boolean apply(ClusterNode node, Message msg) { + if ((msg instanceof GridDhtTxPrepareRequest && prepareBlock.get()) || + ((msg instanceof GridDhtTxFinishRequest || msg instanceof GridDhtAtomicSingleUpdateRequest) && finishBlock.get())) { + CountDownLatch latch = blockLatch.get(); + + assertTrue(latch.getCount() > 0); + + latch.countDown(); + + return true; // Generating counter misses. + } + else + return false; + } + }); + + int blockedFrom = reuseKeys ? 0 : preloadCnt; + int committedFrom = blockedFrom + 1_000; + + int blockedKey = blockedFrom - 1; // None + int committedKey = committedFrom - 1; // None + + List<String> backupMissed = new ArrayList<>(); + List<String> primaryMissed = new ArrayList<>(); + + IgniteCache<Integer, Integer> primCache = prim.cache(DEFAULT_CACHE_NAME); + + Consumer<Integer> cachePut = (key) -> primCache.put(key, -key); + + Set<Thread> ths = ConcurrentHashMap.newKeySet(); + + Consumer<Integer> cachePutAsync = (key) -> { + Thread th = new Thread(() -> cachePut.accept(key)); + + ths.add(th); + + th.start(); + }; + + int primaryLwm = -1; // Primary LWM after data prepatation (at tx caches). + int backupHwm = -1; // Backups HWM after data preparation (at tx caches). + + String backupMissedTail = null; // Misses after backupHwm, which backups are not aware of before the recovery. + + int primaryKeysCnt = preloadCnt; // Keys present on primary. + + int iters = 11; + + ThreadLocalRandom rnd = ThreadLocalRandom.current(); + + // The main idea of this section is to generate missed counters updates, some of them should be on backups only, + // some at primary too. + // This emulates the situation of reordering possible on real highloaded clusters. + for (int it = 0; it < iters; it++) { + boolean first = it == 0; + boolean last = it == iters - 1; + + boolean globalBlock = it % 2 != 0 && // Global means all nodes (including primary) will miss the counters updates. + // Since atomic cache commits entry on primary before sending the request to backups, so all misses are only on backups. + atomicityMode != ATOMIC; + + if (first || last) // Odd number to gain misses on backups only at first and last iteration. + // First iteration will guarantee rebalance for tx caches: Primary LWM > backup LWM. + // Last iteration will guarantee rebalance for atomic caches: Primary counter > backup counter. + assertTrue(!globalBlock); + + int range = rnd.nextInt(1, 5); + + blockLatch.set(new CountDownLatch(backupNodes * range)); + + if (globalBlock) + prepareBlock.set(true); + else + finishBlock.set(true); + + for (int i = 0; i < range; i++) { + blockedKey++; + updateCnt++; + + if (!globalBlock) + primaryKeysCnt++; + + for (Ignite backup : backups) // Check before put. + assertEquals(reuseKeys, backup.cache(DEFAULT_CACHE_NAME).get(blockedKey) != null); + + cachePutAsync.accept(blockedKey); + } + + blockLatch.get().await(); + + prepareBlock.set(false); + finishBlock.set(false); + + String missed = range == 1 ? String.valueOf(updateCnt) : (updateCnt - range + 1) + " - " + updateCnt; + + if (last) + backupMissedTail = missed; + else + backupMissed.add(missed); + + if (globalBlock) + primaryMissed.add(missed); + + if (!last) + for (int i = 0; i < range; i++) { + committedKey++; + updateCnt++; + primaryKeysCnt++; + + cachePut.accept(committedKey); + } + + if (first) + primaryLwm = updateCnt; + + if (!last) + backupHwm = updateCnt; + } + + assertNotNull(backupMissedTail); + assertTrue(primaryLwm != -1); + assertTrue(backupHwm != -1); + assertTrue(blockedKey < committedFrom); // Intersection check. + + if (reuseKeys) + assertTrue(blockedKey < preloadCnt); // Owerflow check. + + for (int key = blockedFrom; key <= blockedKey; key++) { + for (Ignite backup : backups) // Check after put. + assertEquals(reuseKeys, backup.cache(DEFAULT_CACHE_NAME).get(key) != null); + } + + for (int key = committedFrom; key <= committedKey; key++) { + assertNotNull(primCache.get(key)); + + for (Ignite backup : backups) + assertNotNull(backup.cache(DEFAULT_CACHE_NAME).get(key)); + } + + injectTestSystemOut(); + + assertEquals(EXIT_CODE_OK, execute("--cache", "idle_verify")); + + assertConflicts(true, true); + + if (atomicityMode == TRANSACTIONAL) { + assertTxCounters(primaryLwm, primaryMissed, updateCnt); // Primary + assertTxCounters(preloadCnt, backupMissed, backupHwm); // Backups + } + else { + assertAtomicCounters(updateCnt); // Primary + assertAtomicCounters(backupHwm); // Backups + } + + LogListener lsnrRebalanceType = matches("fullPartitions=[" + (historical ? "" : 0) + "], " + + "histPartitions=[" + (historical ? 0 : "") + "]").times(backupNodes).build(); + + LogListener lsnrRebalanceAmount = matches("receivedKeys=" + + (historical ? + atomicityMode == TRANSACTIONAL ? + primaryLwm - preloadCnt : // Diff between primary LWM and backup LWM. + updateCnt - backupHwm + // Diff between primary and backup counters + DFLT_WAL_MARGIN_FOR_ATOMIC_CACHE_HISTORICAL_REBALANCE : // + magic number. + reuseKeys ? + preloadCnt : // Since keys are reused, amount is the same as at initial preloading. + atomicityMode == TRANSACTIONAL ? + primaryKeysCnt : // Calculated amount of entries (initial preload amount + updates not missed at primary) + updateCnt)) // All updates since only misses on backups generated for atomic caches at this test. + .times(backupNodes).build(); + + listeningLog.registerListener(lsnrRebalanceType); + listeningLog.registerListener(lsnrRebalanceAmount); + + ioBlocked = true; // Emulating power off, OOM or disk overflow. Keeping data as is, with missed counters updates. + + stopAllGrids(); + + for (Thread th : ths) + th.join(); + + ioBlocked = false; + + ignite = startGrids(nodes); + + awaitPartitionMapExchange(); + + assertTrue(lsnrRebalanceType.check()); + + assertEquals(EXIT_CODE_OK, execute("--cache", "idle_verify")); + + assertConflicts( + atomicityMode == TRANSACTIONAL, // Atomic cache backup counters are automatically set as primary on crash recovery. + historical); // Full rebalance fixes the consistency. + + assertTrue(lsnrRebalanceAmount.check()); + + if (atomicityMode == TRANSACTIONAL) { // Same as before the crash. + assertTxCounters(primaryLwm, primaryMissed, updateCnt); // Primary + assertTxCounters(preloadCnt, backupMissed, backupHwm); // Backups + } + else if (historical) { + assertAtomicCounters(updateCnt); // Primary + + // Backups updated with primary counter. + // Entries between primary and backup counters are rebalanced. + assertNoneAtomicCounters(backupHwm); // Automaticaly "repaired". + } + else + assertNoneAtomicCounters(); + + assertEquals(EXIT_CODE_OK, execute("--consistency", "repair", + ConsistencyCommand.CACHE, DEFAULT_CACHE_NAME, + ConsistencyCommand.PARTITION, "0", + ConsistencyCommand.STRATEGY, strategy.toString())); + + int repairedCnt = repairedEntriesCount(); + + assertContains(log, testOut.toString(), historical ? CONSISTENCY_VIOLATIONS_FOUND : NOTHING_FOUND); + + assertEquals(EXIT_CODE_OK, execute("--cache", "idle_verify")); + + switch (strategy) { + case PRIMARY: + case REMOVE: + case RELATIVE_MAJORITY: + assertConflicts(atomicityMode == TRANSACTIONAL, false); // Inconsistency fixed. + + break; + + case LWW: + // LWM can't fix the data when some nodes contain the key, but some not, + // because missed entry has no version and can not be compared. + assertConflicts(atomicityMode == TRANSACTIONAL, historical && !reuseKeys); + + break; + + case CHECK_ONLY: // Keeps as is. + assertConflicts(atomicityMode == TRANSACTIONAL, historical); + + break; + + default: + throw new UnsupportedOperationException("Unsupported strategy"); + } + + int postloadFrom = reuseKeys ? 0 : (committedKey + 1 /*greater than max key used*/); + + for (int i = postloadFrom; i < postloadFrom + preloadCnt; i++) { + updateCnt++; + + ignite.cache(DEFAULT_CACHE_NAME).put(i, i); + } + + // Repairing one more time, but with guarantee to fix (primary strategy); + assertEquals(EXIT_CODE_OK, execute("--consistency", "repair", + ConsistencyCommand.CACHE, DEFAULT_CACHE_NAME, + ConsistencyCommand.PARTITION, "0", + ConsistencyCommand.STRATEGY, PRIMARY.toString())); + + repairedCnt += repairedEntriesCount(); + + updateCnt += repairedCnt; + + assertEquals(EXIT_CODE_OK, execute("--cache", "idle_verify")); + + backupMissed.add(backupMissedTail); // Now backups got updates after the generated data and aware of all missed updates. + + if (atomicityMode == TRANSACTIONAL) { + assertConflicts(true, false); // Still has counters conflicts, while inconsistency already fixed. + + assertTxCounters(primaryLwm, primaryMissed, updateCnt); // Primary (same as before crash). + assertTxCounters(preloadCnt, backupMissed, updateCnt); // Backups (all missed updates, increased hwm). + } + else + assertConflicts(false, false); // Fixed both counters and inconsistency. + + int rmvd = strategy == RELATIVE_MAJORITY || strategy == REMOVE ? repairedCnt /*entries removed during the repair*/ : 0; + + for (Ignite node : G.allGrids()) // Checking the cache size after the fixes. + assertEquals((reuseKeys ? preloadCnt : primaryKeysCnt + preloadCnt /*postload*/ - rmvd), + node.cache(DEFAULT_CACHE_NAME).localSize(CachePeekMode.ALL)); + } + + /** + * Checks idle_vefify result. + * + * @param counter Counter conflicts. + * @param hash Hash conflicts. + */ + private void assertConflicts(boolean counter, boolean hash) { + if (counter || hash) + assertContains(log, testOut.toString(), "conflict partitions has been found: " + + "[counterConflicts=" + (counter ? 1 : 0) + ", hashConflicts=" + (hash ? 1 : 0) + "]"); + else + assertContains(log, testOut.toString(), "no conflicts have been found"); + } + + /** + * @param lwm Lwm. + * @param missed Missed. + * @param hwm Hwm. + */ + private void assertTxCounters(int lwm, List<String> missed, int hwm) { + assertContains(log, testOut.toString(), + "updateCntr=[lwm=" + lwm + ", missed=" + missed + ", hwm=" + hwm + "]"); + } + + /** + * @param cnt Counter. + */ + private void assertAtomicCounters(int cnt) { + assertContains(log, testOut.toString(), "updateCntr=" + cnt); + } + + /** + * @param cnt Counter. + */ + private void assertNoneAtomicCounters(int... cnt) { + assertNotContains(log, testOut.toString(), "updateCntr=" + (F.isEmpty(cnt) ? "" : cnt[0])); + } + + /** + * Returns amount of entries repaired by Consistency Repair operation. + */ + private int repairedEntriesCount() { + Pattern pattern = Pattern.compile("repaired=(\\d*),"); + Matcher matcher = pattern.matcher(testOut.toString()); + + if (matcher.find()) + return Integer.parseInt(testOut.toString().substring(matcher.start(1), matcher.end(1))); + else + return 0; Review Comment: Done -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
