Search in sources :

Example 6 with JettySolrRunner

use of org.apache.solr.client.solrj.embedded.JettySolrRunner in project lucene-solr by apache.

the class HttpPartitionTest method testLeaderZkSessionLoss.

// test inspired by SOLR-6511
protected void testLeaderZkSessionLoss() throws Exception {
    String testCollectionName = "c8n_1x2_leader_session_loss";
    createCollectionRetry(testCollectionName, 1, 2, 1);
    cloudClient.setDefaultCollection(testCollectionName);
    sendDoc(1);
    List<Replica> notLeaders = ensureAllReplicasAreActive(testCollectionName, "shard1", 1, 2, maxWaitSecsToSeeAllActive);
    assertTrue("Expected 1 replicas for collection " + testCollectionName + " but found " + notLeaders.size() + "; clusterState: " + printClusterStateInfo(testCollectionName), notLeaders.size() == 1);
    Replica leader = cloudClient.getZkStateReader().getLeaderRetry(testCollectionName, "shard1");
    String leaderNode = leader.getNodeName();
    assertNotNull("Could not find leader for shard1 of " + testCollectionName + "; clusterState: " + printClusterStateInfo(testCollectionName), leader);
    JettySolrRunner leaderJetty = getJettyOnPort(getReplicaPort(leader));
    SolrInputDocument doc = new SolrInputDocument();
    doc.addField(id, String.valueOf(2));
    doc.addField("a_t", "hello" + 2);
    // cause leader migration by expiring the current leader's zk session
    chaosMonkey.expireSession(leaderJetty);
    String expectedNewLeaderCoreNodeName = notLeaders.get(0).getName();
    long timeout = System.nanoTime() + TimeUnit.NANOSECONDS.convert(60, TimeUnit.SECONDS);
    while (System.nanoTime() < timeout) {
        String currentLeaderName = null;
        try {
            Replica currentLeader = cloudClient.getZkStateReader().getLeaderRetry(testCollectionName, "shard1");
            currentLeaderName = currentLeader.getName();
        } catch (Exception exc) {
        }
        if (expectedNewLeaderCoreNodeName.equals(currentLeaderName))
            // new leader was elected after zk session expiration
            break;
        Thread.sleep(500);
    }
    Replica currentLeader = cloudClient.getZkStateReader().getLeaderRetry(testCollectionName, "shard1");
    assertEquals(expectedNewLeaderCoreNodeName, currentLeader.getName());
    // TODO: This test logic seems to be timing dependent and fails on Jenkins
    // need to come up with a better approach
    log.info("Sending doc 2 to old leader " + leader.getName());
    try (HttpSolrClient leaderSolr = getHttpSolrClient(leader, testCollectionName)) {
        leaderSolr.add(doc);
        leaderSolr.close();
        // if the add worked, then the doc must exist on the new leader
        try (HttpSolrClient newLeaderSolr = getHttpSolrClient(currentLeader, testCollectionName)) {
            assertDocExists(newLeaderSolr, testCollectionName, "2");
        }
    } catch (SolrException exc) {
        // this is ok provided the doc doesn't exist on the current leader
        try (HttpSolrClient client = getHttpSolrClient(currentLeader, testCollectionName)) {
            // this should work
            client.add(doc);
        }
    }
    List<Replica> participatingReplicas = getActiveOrRecoveringReplicas(testCollectionName, "shard1");
    Set<String> replicasToCheck = new HashSet<>();
    for (Replica stillUp : participatingReplicas) replicasToCheck.add(stillUp.getName());
    waitToSeeReplicasActive(testCollectionName, "shard1", replicasToCheck, 20);
    assertDocsExistInAllReplicas(participatingReplicas, testCollectionName, 1, 2);
    log.info("testLeaderZkSessionLoss succeeded ... deleting the " + testCollectionName + " collection");
    // try to clean up
    attemptCollectionDelete(cloudClient, testCollectionName);
}
Also used : HttpSolrClient(org.apache.solr.client.solrj.impl.HttpSolrClient) SolrInputDocument(org.apache.solr.common.SolrInputDocument) JettySolrRunner(org.apache.solr.client.solrj.embedded.JettySolrRunner) Replica(org.apache.solr.common.cloud.Replica) SolrServerException(org.apache.solr.client.solrj.SolrServerException) SolrException(org.apache.solr.common.SolrException) IOException(java.io.IOException) SolrException(org.apache.solr.common.SolrException) HashSet(java.util.HashSet)

Example 7 with JettySolrRunner

use of org.apache.solr.client.solrj.embedded.JettySolrRunner in project lucene-solr by apache.

the class HttpPartitionTest method testLeaderInitiatedRecoveryCRUD.

/**
   * Tests handling of lir state znodes.
   */
protected void testLeaderInitiatedRecoveryCRUD() throws Exception {
    String testCollectionName = "c8n_crud_1x2";
    String shardId = "shard1";
    createCollectionRetry(testCollectionName, 1, 2, 1);
    cloudClient.setDefaultCollection(testCollectionName);
    Replica leader = cloudClient.getZkStateReader().getLeaderRetry(testCollectionName, shardId);
    JettySolrRunner leaderJetty = getJettyOnPort(getReplicaPort(leader));
    CoreContainer cores = leaderJetty.getCoreContainer();
    ZkController zkController = cores.getZkController();
    assertNotNull("ZkController is null", zkController);
    Replica notLeader = ensureAllReplicasAreActive(testCollectionName, shardId, 1, 2, maxWaitSecsToSeeAllActive).get(0);
    ZkCoreNodeProps replicaCoreNodeProps = new ZkCoreNodeProps(notLeader);
    String replicaUrl = replicaCoreNodeProps.getCoreUrl();
    MockCoreDescriptor cd = new MockCoreDescriptor() {

        public CloudDescriptor getCloudDescriptor() {
            return new CloudDescriptor(leader.getStr(ZkStateReader.CORE_NAME_PROP), new Properties(), this) {

                @Override
                public String getCoreNodeName() {
                    return leader.getName();
                }

                @Override
                public boolean isLeader() {
                    return true;
                }
            };
        }
    };
    zkController.updateLeaderInitiatedRecoveryState(testCollectionName, shardId, notLeader.getName(), Replica.State.DOWN, cd, true);
    Map<String, Object> lirStateMap = zkController.getLeaderInitiatedRecoveryStateObject(testCollectionName, shardId, notLeader.getName());
    assertNotNull(lirStateMap);
    assertSame(Replica.State.DOWN, Replica.State.getState((String) lirStateMap.get(ZkStateReader.STATE_PROP)));
    // test old non-json format handling
    SolrZkClient zkClient = zkController.getZkClient();
    String znodePath = zkController.getLeaderInitiatedRecoveryZnodePath(testCollectionName, shardId, notLeader.getName());
    zkClient.setData(znodePath, "down".getBytes(StandardCharsets.UTF_8), true);
    lirStateMap = zkController.getLeaderInitiatedRecoveryStateObject(testCollectionName, shardId, notLeader.getName());
    assertNotNull(lirStateMap);
    assertSame(Replica.State.DOWN, Replica.State.getState((String) lirStateMap.get(ZkStateReader.STATE_PROP)));
    zkClient.delete(znodePath, -1, false);
    // try to clean up
    attemptCollectionDelete(cloudClient, testCollectionName);
}
Also used : CoreContainer(org.apache.solr.core.CoreContainer) ZkCoreNodeProps(org.apache.solr.common.cloud.ZkCoreNodeProps) JettySolrRunner(org.apache.solr.client.solrj.embedded.JettySolrRunner) MockCoreDescriptor(org.apache.solr.util.MockCoreContainer.MockCoreDescriptor) Properties(java.util.Properties) Replica(org.apache.solr.common.cloud.Replica) SolrZkClient(org.apache.solr.common.cloud.SolrZkClient)

Example 8 with JettySolrRunner

use of org.apache.solr.client.solrj.embedded.JettySolrRunner in project lucene-solr by apache.

the class TestCloudRecovery method leaderRecoverFromLogOnStartupTest.

@Test
public void leaderRecoverFromLogOnStartupTest() throws Exception {
    AtomicInteger countReplayLog = new AtomicInteger(0);
    DirectUpdateHandler2.commitOnClose = false;
    UpdateLog.testing_logReplayFinishHook = countReplayLog::incrementAndGet;
    CloudSolrClient cloudClient = cluster.getSolrClient();
    cloudClient.add(COLLECTION, sdoc("id", "1"));
    cloudClient.add(COLLECTION, sdoc("id", "2"));
    cloudClient.add(COLLECTION, sdoc("id", "3"));
    cloudClient.add(COLLECTION, sdoc("id", "4"));
    ModifiableSolrParams params = new ModifiableSolrParams();
    params.set("q", "*:*");
    QueryResponse resp = cloudClient.query(COLLECTION, params);
    assertEquals(0, resp.getResults().getNumFound());
    ChaosMonkey.stop(cluster.getJettySolrRunners());
    assertTrue("Timeout waiting for all not live", ClusterStateUtil.waitForAllReplicasNotLive(cloudClient.getZkStateReader(), 45000));
    ChaosMonkey.start(cluster.getJettySolrRunners());
    assertTrue("Timeout waiting for all live and active", ClusterStateUtil.waitForAllActiveAndLiveReplicas(cloudClient.getZkStateReader(), COLLECTION, 120000));
    resp = cloudClient.query(COLLECTION, params);
    assertEquals(4, resp.getResults().getNumFound());
    // Make sure all nodes is recover from tlog
    if (onlyLeaderIndexes) {
        // Leader election can be kicked off, so 2 tlog replicas will replay its tlog before becoming new leader
        assertTrue(countReplayLog.get() >= 2);
    } else {
        assertEquals(4, countReplayLog.get());
    }
    // check metrics
    int replicationCount = 0;
    int errorsCount = 0;
    int skippedCount = 0;
    for (JettySolrRunner jetty : cluster.getJettySolrRunners()) {
        SolrMetricManager manager = jetty.getCoreContainer().getMetricManager();
        List<String> registryNames = manager.registryNames().stream().filter(s -> s.startsWith("solr.core.")).collect(Collectors.toList());
        for (String registry : registryNames) {
            Map<String, Metric> metrics = manager.registry(registry).getMetrics();
            Timer timer = (Timer) metrics.get("REPLICATION.peerSync.time");
            Counter counter = (Counter) metrics.get("REPLICATION.peerSync.errors");
            Counter skipped = (Counter) metrics.get("REPLICATION.peerSync.skipped");
            replicationCount += timer.getCount();
            errorsCount += counter.getCount();
            skippedCount += skipped.getCount();
        }
    }
    if (onlyLeaderIndexes) {
        assertTrue(replicationCount >= 2);
    } else {
        assertEquals(2, replicationCount);
    }
}
Also used : Arrays(java.util.Arrays) BeforeClass(org.junit.BeforeClass) HashMap(java.util.HashMap) DirectUpdateHandler2(org.apache.solr.update.DirectUpdateHandler2) SolrServerException(org.apache.solr.client.solrj.SolrServerException) ClusterStateUtil(org.apache.solr.common.cloud.ClusterStateUtil) AtomicInteger(java.util.concurrent.atomic.AtomicInteger) Map(java.util.Map) Counter(com.codahale.metrics.Counter) UpdateLog(org.apache.solr.update.UpdateLog) JettySolrRunner(org.apache.solr.client.solrj.embedded.JettySolrRunner) CloudSolrClient(org.apache.solr.client.solrj.impl.CloudSolrClient) Before(org.junit.Before) AfterClass(org.junit.AfterClass) ModifiableSolrParams(org.apache.solr.common.params.ModifiableSolrParams) SolrCore(org.apache.solr.core.SolrCore) QueryResponse(org.apache.solr.client.solrj.response.QueryResponse) FileOutputStream(java.io.FileOutputStream) Metric(com.codahale.metrics.Metric) IOException(java.io.IOException) Test(org.junit.Test) FileInputStream(java.io.FileInputStream) Collectors(java.util.stream.Collectors) File(java.io.File) IOUtils(org.apache.commons.io.IOUtils) List(java.util.List) Timer(com.codahale.metrics.Timer) SolrMetricManager(org.apache.solr.metrics.SolrMetricManager) TestInjection(org.apache.solr.util.TestInjection) CollectionAdminRequest(org.apache.solr.client.solrj.request.CollectionAdminRequest) JettySolrRunner(org.apache.solr.client.solrj.embedded.JettySolrRunner) ModifiableSolrParams(org.apache.solr.common.params.ModifiableSolrParams) CloudSolrClient(org.apache.solr.client.solrj.impl.CloudSolrClient) Counter(com.codahale.metrics.Counter) Timer(com.codahale.metrics.Timer) AtomicInteger(java.util.concurrent.atomic.AtomicInteger) QueryResponse(org.apache.solr.client.solrj.response.QueryResponse) SolrMetricManager(org.apache.solr.metrics.SolrMetricManager) Metric(com.codahale.metrics.Metric) Test(org.junit.Test)

Example 9 with JettySolrRunner

use of org.apache.solr.client.solrj.embedded.JettySolrRunner in project lucene-solr by apache.

the class TestCloudRecovery method corruptedLogTest.

@Test
public void corruptedLogTest() throws Exception {
    AtomicInteger countReplayLog = new AtomicInteger(0);
    DirectUpdateHandler2.commitOnClose = false;
    UpdateLog.testing_logReplayFinishHook = countReplayLog::incrementAndGet;
    CloudSolrClient cloudClient = cluster.getSolrClient();
    cloudClient.add(COLLECTION, sdoc("id", "1000"));
    cloudClient.add(COLLECTION, sdoc("id", "1001"));
    for (int i = 0; i < 10; i++) {
        cloudClient.add(COLLECTION, sdoc("id", String.valueOf(i)));
    }
    ModifiableSolrParams params = new ModifiableSolrParams();
    params.set("q", "*:*");
    QueryResponse resp = cloudClient.query(COLLECTION, params);
    assertEquals(0, resp.getResults().getNumFound());
    int logHeaderSize = Integer.MAX_VALUE;
    Map<File, byte[]> contentFiles = new HashMap<>();
    for (JettySolrRunner solrRunner : cluster.getJettySolrRunners()) {
        for (SolrCore solrCore : solrRunner.getCoreContainer().getCores()) {
            File tlogFolder = new File(solrCore.getUlogDir(), UpdateLog.TLOG_NAME);
            String[] tLogFiles = tlogFolder.list();
            Arrays.sort(tLogFiles);
            File lastTLogFile = new File(tlogFolder.getAbsolutePath() + "/" + tLogFiles[tLogFiles.length - 1]);
            byte[] tlogBytes = IOUtils.toByteArray(new FileInputStream(lastTLogFile));
            contentFiles.put(lastTLogFile, tlogBytes);
            logHeaderSize = Math.min(tlogBytes.length, logHeaderSize);
        }
    }
    ChaosMonkey.stop(cluster.getJettySolrRunners());
    assertTrue("Timeout waiting for all not live", ClusterStateUtil.waitForAllReplicasNotLive(cloudClient.getZkStateReader(), 45000));
    for (Map.Entry<File, byte[]> entry : contentFiles.entrySet()) {
        byte[] tlogBytes = entry.getValue();
        if (tlogBytes.length <= logHeaderSize)
            continue;
        FileOutputStream stream = new FileOutputStream(entry.getKey());
        int skipLastBytes = Math.max(random().nextInt(tlogBytes.length - logHeaderSize), 2);
        for (int i = 0; i < entry.getValue().length - skipLastBytes; i++) {
            stream.write(tlogBytes[i]);
        }
        stream.close();
    }
    ChaosMonkey.start(cluster.getJettySolrRunners());
    assertTrue("Timeout waiting for all live and active", ClusterStateUtil.waitForAllActiveAndLiveReplicas(cloudClient.getZkStateReader(), COLLECTION, 120000));
    resp = cloudClient.query(COLLECTION, params);
    // Make sure cluster still healthy
    assertTrue(resp.getResults().getNumFound() >= 2);
}
Also used : HashMap(java.util.HashMap) JettySolrRunner(org.apache.solr.client.solrj.embedded.JettySolrRunner) SolrCore(org.apache.solr.core.SolrCore) ModifiableSolrParams(org.apache.solr.common.params.ModifiableSolrParams) FileInputStream(java.io.FileInputStream) CloudSolrClient(org.apache.solr.client.solrj.impl.CloudSolrClient) AtomicInteger(java.util.concurrent.atomic.AtomicInteger) QueryResponse(org.apache.solr.client.solrj.response.QueryResponse) FileOutputStream(java.io.FileOutputStream) File(java.io.File) HashMap(java.util.HashMap) Map(java.util.Map) Test(org.junit.Test)

Example 10 with JettySolrRunner

use of org.apache.solr.client.solrj.embedded.JettySolrRunner in project lucene-solr by apache.

the class TestMiniSolrCloudCluster method testStopAllStartAll.

@Test
public void testStopAllStartAll() throws Exception {
    final String collectionName = "testStopAllStartAllCollection";
    final MiniSolrCloudCluster miniCluster = createMiniSolrCloudCluster();
    try {
        assertNotNull(miniCluster.getZkServer());
        List<JettySolrRunner> jettys = miniCluster.getJettySolrRunners();
        assertEquals(NUM_SERVERS, jettys.size());
        for (JettySolrRunner jetty : jettys) {
            assertTrue(jetty.isRunning());
        }
        createCollection(miniCluster, collectionName, null, null, Boolean.TRUE, null);
        final CloudSolrClient cloudSolrClient = miniCluster.getSolrClient();
        cloudSolrClient.setDefaultCollection(collectionName);
        final SolrQuery query = new SolrQuery("*:*");
        final SolrInputDocument doc = new SolrInputDocument();
        try (SolrZkClient zkClient = new SolrZkClient(miniCluster.getZkServer().getZkAddress(), AbstractZkTestCase.TIMEOUT, AbstractZkTestCase.TIMEOUT, null);
            ZkStateReader zkStateReader = new ZkStateReader(zkClient)) {
            zkStateReader.createClusterStateWatchersAndUpdate();
            AbstractDistribZkTestBase.waitForRecoveriesToFinish(collectionName, zkStateReader, true, true, 330);
            // modify collection
            final int numDocs = 1 + random().nextInt(10);
            for (int ii = 1; ii <= numDocs; ++ii) {
                doc.setField("id", "" + ii);
                cloudSolrClient.add(doc);
                if (ii * 2 == numDocs)
                    cloudSolrClient.commit();
            }
            cloudSolrClient.commit();
            // query collection
            {
                final QueryResponse rsp = cloudSolrClient.query(query);
                assertEquals(numDocs, rsp.getResults().getNumFound());
            }
            // the test itself
            zkStateReader.forceUpdateCollection(collectionName);
            final ClusterState clusterState = zkStateReader.getClusterState();
            final HashSet<Integer> leaderIndices = new HashSet<Integer>();
            final HashSet<Integer> followerIndices = new HashSet<Integer>();
            {
                final HashMap<String, Boolean> shardLeaderMap = new HashMap<String, Boolean>();
                for (final Slice slice : clusterState.getSlices(collectionName)) {
                    for (final Replica replica : slice.getReplicas()) {
                        shardLeaderMap.put(replica.getNodeName().replace("_solr", "/solr"), Boolean.FALSE);
                    }
                    shardLeaderMap.put(slice.getLeader().getNodeName().replace("_solr", "/solr"), Boolean.TRUE);
                }
                for (int ii = 0; ii < jettys.size(); ++ii) {
                    final URL jettyBaseUrl = jettys.get(ii).getBaseUrl();
                    final String jettyBaseUrlString = jettyBaseUrl.toString().substring((jettyBaseUrl.getProtocol() + "://").length());
                    final Boolean isLeader = shardLeaderMap.get(jettyBaseUrlString);
                    if (Boolean.TRUE.equals(isLeader)) {
                        leaderIndices.add(new Integer(ii));
                    } else if (Boolean.FALSE.equals(isLeader)) {
                        followerIndices.add(new Integer(ii));
                    }
                // else neither leader nor follower i.e. node without a replica (for our collection)
                }
            }
            final List<Integer> leaderIndicesList = new ArrayList<Integer>(leaderIndices);
            final List<Integer> followerIndicesList = new ArrayList<Integer>(followerIndices);
            // first stop the followers (in no particular order)
            Collections.shuffle(followerIndicesList, random());
            for (Integer ii : followerIndicesList) {
                if (!leaderIndices.contains(ii)) {
                    miniCluster.stopJettySolrRunner(jettys.get(ii.intValue()));
                }
            }
            // then stop the leaders (again in no particular order)
            Collections.shuffle(leaderIndicesList, random());
            for (Integer ii : leaderIndicesList) {
                miniCluster.stopJettySolrRunner(jettys.get(ii.intValue()));
            }
            // calculate restart order
            final List<Integer> restartIndicesList = new ArrayList<Integer>();
            Collections.shuffle(leaderIndicesList, random());
            restartIndicesList.addAll(leaderIndicesList);
            Collections.shuffle(followerIndicesList, random());
            restartIndicesList.addAll(followerIndicesList);
            if (random().nextBoolean())
                Collections.shuffle(restartIndicesList, random());
            // and then restart jettys in that order
            for (Integer ii : restartIndicesList) {
                final JettySolrRunner jetty = jettys.get(ii.intValue());
                if (!jetty.isRunning()) {
                    miniCluster.startJettySolrRunner(jetty);
                    assertTrue(jetty.isRunning());
                }
            }
            AbstractDistribZkTestBase.waitForRecoveriesToFinish(collectionName, zkStateReader, true, true, 330);
            zkStateReader.forceUpdateCollection(collectionName);
            // re-query collection
            {
                final QueryResponse rsp = cloudSolrClient.query(query);
                assertEquals(numDocs, rsp.getResults().getNumFound());
            }
        }
    } finally {
        miniCluster.shutdown();
    }
}
Also used : ClusterState(org.apache.solr.common.cloud.ClusterState) HashMap(java.util.HashMap) JettySolrRunner(org.apache.solr.client.solrj.embedded.JettySolrRunner) ArrayList(java.util.ArrayList) SolrZkClient(org.apache.solr.common.cloud.SolrZkClient) Replica(org.apache.solr.common.cloud.Replica) SolrQuery(org.apache.solr.client.solrj.SolrQuery) URL(java.net.URL) CloudSolrClient(org.apache.solr.client.solrj.impl.CloudSolrClient) ZkStateReader(org.apache.solr.common.cloud.ZkStateReader) SolrInputDocument(org.apache.solr.common.SolrInputDocument) Slice(org.apache.solr.common.cloud.Slice) QueryResponse(org.apache.solr.client.solrj.response.QueryResponse) HashSet(java.util.HashSet) Test(org.junit.Test)

Aggregations

JettySolrRunner (org.apache.solr.client.solrj.embedded.JettySolrRunner)137 Test (org.junit.Test)52 HttpSolrClient (org.apache.solr.client.solrj.impl.HttpSolrClient)29 Replica (org.apache.solr.common.cloud.Replica)28 CloudSolrClient (org.apache.solr.client.solrj.impl.CloudSolrClient)25 SolrInputDocument (org.apache.solr.common.SolrInputDocument)20 ModifiableSolrParams (org.apache.solr.common.params.ModifiableSolrParams)20 File (java.io.File)19 SolrQuery (org.apache.solr.client.solrj.SolrQuery)19 DocCollection (org.apache.solr.common.cloud.DocCollection)18 Slice (org.apache.solr.common.cloud.Slice)18 IOException (java.io.IOException)15 ArrayList (java.util.ArrayList)15 Properties (java.util.Properties)15 SolrClient (org.apache.solr.client.solrj.SolrClient)15 UpdateRequest (org.apache.solr.client.solrj.request.UpdateRequest)15 QueryResponse (org.apache.solr.client.solrj.response.QueryResponse)15 CoreContainer (org.apache.solr.core.CoreContainer)14 BeforeClass (org.junit.BeforeClass)14 ClusterState (org.apache.solr.common.cloud.ClusterState)13