Examples with Replica org.apache.solr.common.cloud.Replica used on opensource projects

Example 46 with Replica

use of org.apache.solr.common.cloud.Replica in project lucene-solr by apache.

the class ReplicaPropertiesBase method verifyPropertyNotPresent.

public static void verifyPropertyNotPresent(CloudSolrClient client, String collectionName, String replicaName, String property) throws KeeperException, InterruptedException {
    ClusterState clusterState = null;
    Replica replica = null;
    for (int idx = 0; idx < 300; ++idx) {
        clusterState = client.getZkStateReader().getClusterState();
        replica = clusterState.getReplica(collectionName, replicaName);
        if (replica == null) {
            fail("Could not find collection/replica pair! " + collectionName + "/" + replicaName);
        }
        if (StringUtils.isBlank(replica.getStr(property)))
            return;
        Thread.sleep(100);
    }
    fail("Property " + property + " not set correctly for collection/replica pair: " + collectionName + "/" + replicaName + ". Replica props: " + replica.getProperties().toString() + ". Cluster state is " + clusterState.toString());
}

Example 47 with Replica

use of org.apache.solr.common.cloud.Replica in project lucene-solr by apache.

the class ReplicationFactorTest method testRf3.

protected void testRf3() throws Exception {
    int numShards = 1;
    int replicationFactor = 3;
    int maxShardsPerNode = 1;
    String testCollectionName = "repfacttest_c8n_1x3";
    String shardId = "shard1";
    int minRf = 2;
    createCollection(testCollectionName, numShards, replicationFactor, maxShardsPerNode);
    cloudClient.setDefaultCollection(testCollectionName);
    List<Replica> replicas = ensureAllReplicasAreActive(testCollectionName, shardId, numShards, replicationFactor, 30);
    assertTrue("Expected 2 active replicas for " + testCollectionName, replicas.size() == 2);
    int rf = sendDoc(1, minRf);
    assertRf(3, "all replicas should be active", rf);
    getProxyForReplica(replicas.get(0)).close();
    rf = sendDoc(2, minRf);
    assertRf(2, "one replica should be down", rf);
    getProxyForReplica(replicas.get(1)).close();
    rf = sendDoc(3, minRf);
    assertRf(1, "both replicas should be down", rf);
    // heal the partitions
    getProxyForReplica(replicas.get(0)).reopen();
    getProxyForReplica(replicas.get(1)).reopen();
    // give time for the healed partition to get propagated
    Thread.sleep(2000);
    ensureAllReplicasAreActive(testCollectionName, shardId, numShards, replicationFactor, 30);
    rf = sendDoc(4, minRf);
    assertRf(3, "partitions to replicas have been healed", rf);
    // now send a batch
    List<SolrInputDocument> batch = new ArrayList<SolrInputDocument>(10);
    for (int i = 5; i < 15; i++) {
        SolrInputDocument doc = new SolrInputDocument();
        doc.addField(id, String.valueOf(i));
        doc.addField("a_t", "hello" + i);
        batch.add(doc);
    }
    int batchRf = sendDocsWithRetry(batch, minRf, 5, 1);
    assertRf(3, "batch should have succeeded on all replicas", batchRf);
    // add some chaos to the batch
    getProxyForReplica(replicas.get(0)).close();
    // now send a batch
    batch = new ArrayList<SolrInputDocument>(10);
    for (int i = 15; i < 30; i++) {
        SolrInputDocument doc = new SolrInputDocument();
        doc.addField(id, String.valueOf(i));
        doc.addField("a_t", "hello" + i);
        batch.add(doc);
    }
    batchRf = sendDocsWithRetry(batch, minRf, 5, 1);
    assertRf(2, "batch should have succeeded on 2 replicas (only one replica should be down)", batchRf);
    // close the 2nd replica, and send a 3rd batch with expected achieved rf=1
    getProxyForReplica(replicas.get(1)).close();
    batch = new ArrayList<SolrInputDocument>(10);
    for (int i = 30; i < 45; i++) {
        SolrInputDocument doc = new SolrInputDocument();
        doc.addField(id, String.valueOf(i));
        doc.addField("a_t", "hello" + i);
        batch.add(doc);
    }
    batchRf = sendDocsWithRetry(batch, minRf, 5, 1);
    assertRf(1, "batch should have succeeded on the leader only (both replicas should be down)", batchRf);
    getProxyForReplica(replicas.get(0)).reopen();
    getProxyForReplica(replicas.get(1)).reopen();
    Thread.sleep(2000);
    ensureAllReplicasAreActive(testCollectionName, shardId, numShards, replicationFactor, 30);
}

Example 48 with Replica

use of org.apache.solr.common.cloud.Replica in project lucene-solr by apache.

the class ForceLeaderTest method setReplicaState.

protected void setReplicaState(String collection, String slice, Replica replica, Replica.State state) throws SolrServerException, IOException, KeeperException, InterruptedException {
    DistributedQueue inQueue = Overseer.getStateUpdateQueue(cloudClient.getZkStateReader().getZkClient());
    ZkStateReader zkStateReader = cloudClient.getZkStateReader();
    String baseUrl = zkStateReader.getBaseUrlForNodeName(replica.getNodeName());
    ZkNodeProps m = new ZkNodeProps(Overseer.QUEUE_OPERATION, OverseerAction.STATE.toLower(), ZkStateReader.BASE_URL_PROP, baseUrl, ZkStateReader.NODE_NAME_PROP, replica.getNodeName(), ZkStateReader.SHARD_ID_PROP, slice, ZkStateReader.COLLECTION_PROP, collection, ZkStateReader.CORE_NAME_PROP, replica.getStr(CORE_NAME_PROP), ZkStateReader.CORE_NODE_NAME_PROP, replica.getName(), ZkStateReader.STATE_PROP, state.toString());
    inQueue.offer(Utils.toJSON(m));
    boolean transition = false;
    Replica.State replicaState = null;
    for (int counter = 10; counter > 0; counter--) {
        ClusterState clusterState = zkStateReader.getClusterState();
        replicaState = clusterState.getSlice(collection, slice).getReplica(replica.getName()).getState();
        if (replicaState == state) {
            transition = true;
            break;
        }
        Thread.sleep(1000);
    }
    if (!transition) {
        throw new SolrException(SolrException.ErrorCode.SERVER_ERROR, "Could not set replica [" + replica.getName() + "] as " + state + ". Last known state of the replica: " + replicaState);
    }
}

Example 49 with Replica

use of org.apache.solr.common.cloud.Replica in project lucene-solr by apache.

the class ForceLeaderTest method testLastPublishedStateIsActive.

/**
   * Test that FORCELEADER can set last published state of all down (live) replicas to active (so
   * that they become worthy candidates for leader election).
   */
@Slow
public void testLastPublishedStateIsActive() throws Exception {
    handle.put("maxScore", SKIPVAL);
    handle.put("timestamp", SKIPVAL);
    String testCollectionName = "forceleader_last_published";
    createCollection(testCollectionName, 1, 3, 1);
    cloudClient.setDefaultCollection(testCollectionName);
    log.info("Collection created: " + testCollectionName);
    try {
        List<Replica> notLeaders = ensureAllReplicasAreActive(testCollectionName, SHARD1, 1, 3, maxWaitSecsToSeeAllActive);
        assertEquals("Expected 2 replicas for collection " + testCollectionName + " but found " + notLeaders.size() + "; clusterState: " + printClusterStateInfo(testCollectionName), 2, notLeaders.size());
        Replica leader = cloudClient.getZkStateReader().getLeaderRetry(testCollectionName, SHARD1);
        JettySolrRunner notLeader0 = getJettyOnPort(getReplicaPort(notLeaders.get(0)));
        ZkController zkController = notLeader0.getCoreContainer().getZkController();
        // Mark all replicas down
        setReplicaState(testCollectionName, SHARD1, leader, State.DOWN);
        for (Replica rep : notLeaders) {
            setReplicaState(testCollectionName, SHARD1, rep, State.DOWN);
        }
        zkController.getZkStateReader().forceUpdateCollection(testCollectionName);
        // Assert all replicas are down and that there is no leader
        assertEquals(0, getActiveOrRecoveringReplicas(testCollectionName, SHARD1).size());
        // Now force leader
        doForceLeader(cloudClient, testCollectionName, SHARD1);
        // Assert that last published states of the two replicas are active now
        for (Replica rep : notLeaders) {
            assertEquals(Replica.State.ACTIVE, getLastPublishedState(testCollectionName, SHARD1, rep));
        }
    } finally {
        log.info("Cleaning up after the test.");
        attemptCollectionDelete(cloudClient, testCollectionName);
    }
}

Example 50 with Replica

use of org.apache.solr.common.cloud.Replica in project lucene-solr by apache.

the class RecoveryAfterSoftCommitTest method test.

@Test
public void test() throws Exception {
    waitForRecoveriesToFinish(DEFAULT_COLLECTION, true);
    // flush twice
    int i = 0;
    for (; i < MAX_BUFFERED_DOCS + 1; i++) {
        SolrInputDocument document = new SolrInputDocument();
        document.addField("id", String.valueOf(i));
        document.addField("a_t", "text_" + i);
        cloudClient.add(document);
    }
    // soft-commit so searchers are open on un-committed but flushed segment files
    AbstractUpdateRequest request = new UpdateRequest().setAction(AbstractUpdateRequest.ACTION.COMMIT, true, true, true);
    cloudClient.request(request);
    Replica notLeader = ensureAllReplicasAreActive(DEFAULT_COLLECTION, "shard1", 1, 2, 30).get(0);
    // ok, now introduce a network partition between the leader and the replica
    SocketProxy proxy = getProxyForReplica(notLeader);
    proxy.close();
    // add more than ULOG_NUM_RECORDS_TO_KEEP docs so that peer sync cannot be used for recovery
    int MAX_DOCS = 2 + MAX_BUFFERED_DOCS + ULOG_NUM_RECORDS_TO_KEEP;
    for (; i < MAX_DOCS; i++) {
        SolrInputDocument document = new SolrInputDocument();
        document.addField("id", String.valueOf(i));
        document.addField("a_t", "text_" + i);
        cloudClient.add(document);
    }
    // Have the partition last at least 1 sec
    // While this gives the impression that recovery is timing related, this is
    // really only
    // to give time for the state to be written to ZK before the test completes.
    // In other words,
    // without a brief pause, the test finishes so quickly that it doesn't give
    // time for the recovery process to kick-in
    Thread.sleep(2000L);
    proxy.reopen();
    List<Replica> notLeaders = ensureAllReplicasAreActive(DEFAULT_COLLECTION, "shard1", 1, 2, 30);
}