Examples with Status org.apache.accumulo.server.replication.proto.Replication.Status used on opensource projects

Example 1 with Status

use of org.apache.accumulo.server.replication.proto.Replication.Status in project accumulo by apache.

the class MockReplicaSystem method replicate.

@Override
public Status replicate(Path p, Status status, ReplicationTarget target, ReplicaSystemHelper helper) {
    Status newStatus;
    if (status.getClosed() && status.getInfiniteEnd()) {
        Status.Builder builder = Status.newBuilder(status);
        if (status.getInfiniteEnd()) {
            builder.setBegin(Long.MAX_VALUE);
        } else {
            builder.setBegin(status.getEnd());
        }
        newStatus = builder.build();
    } else {
        log.info("{} with status {} is not closed and with infinite length, ignoring", p, status);
        newStatus = status;
    }
    log.debug("Sleeping for {}ms before finishing replication on {}", sleep, p);
    try {
        Thread.sleep(sleep);
    } catch (InterruptedException e) {
        log.error("Interrupted while sleeping, will report no progress", e);
        Thread.currentThread().interrupt();
        return status;
    }
    log.info("For {}, received {}, returned {}", p, ProtobufUtil.toString(status), ProtobufUtil.toString(newStatus));
    try {
        helper.recordNewStatus(p, newStatus, target);
    } catch (TableNotFoundException e) {
        log.error("Tried to update status in replication table for {} as {}, but the table did not exist", p, ProtobufUtil.toString(newStatus), e);
        return status;
    } catch (AccumuloException | AccumuloSecurityException e) {
        log.error("Tried to record new status in replication table for {} as {}, but got an error", p, ProtobufUtil.toString(newStatus), e);
        return status;
    }
    return newStatus;
}

Example 2 with Status

use of org.apache.accumulo.server.replication.proto.Replication.Status in project accumulo by apache.

the class MultiInstanceReplicationIT method dataReplicatedToCorrectTableWithoutDrain.

@Test
public void dataReplicatedToCorrectTableWithoutDrain() throws Exception {
    MiniAccumuloConfigImpl peerCfg = new MiniAccumuloConfigImpl(createTestDir(this.getClass().getName() + "_" + this.testName.getMethodName() + "_peer"), ROOT_PASSWORD);
    peerCfg.setNumTservers(1);
    peerCfg.setInstanceName("peer");
    peerCfg.setProperty(Property.REPLICATION_NAME, "peer");
    updatePeerConfigFromPrimary(getCluster().getConfig(), peerCfg);
    MiniAccumuloClusterImpl peer1Cluster = new MiniAccumuloClusterImpl(peerCfg);
    peer1Cluster.start();
    try {
        Connector connMaster = getConnector();
        Connector connPeer = peer1Cluster.getConnector("root", new PasswordToken(ROOT_PASSWORD));
        String peerClusterName = "peer";
        String peerUserName = "repl";
        String peerPassword = "passwd";
        // Create a user on the peer for replication to use
        connPeer.securityOperations().createLocalUser(peerUserName, new PasswordToken(peerPassword));
        // Configure the credentials we should use to authenticate ourselves to the peer for replication
        connMaster.instanceOperations().setProperty(Property.REPLICATION_PEER_USER.getKey() + peerClusterName, peerUserName);
        connMaster.instanceOperations().setProperty(Property.REPLICATION_PEER_PASSWORD.getKey() + peerClusterName, peerPassword);
        // ...peer = AccumuloReplicaSystem,instanceName,zookeepers
        connMaster.instanceOperations().setProperty(Property.REPLICATION_PEERS.getKey() + peerClusterName, ReplicaSystemFactory.getPeerConfigurationValue(AccumuloReplicaSystem.class, AccumuloReplicaSystem.buildConfiguration(peer1Cluster.getInstanceName(), peer1Cluster.getZooKeepers())));
        String masterTable1 = "master1", peerTable1 = "peer1", masterTable2 = "master2", peerTable2 = "peer2";
        connMaster.tableOperations().create(masterTable1);
        String masterTableId1 = connMaster.tableOperations().tableIdMap().get(masterTable1);
        Assert.assertNotNull(masterTableId1);
        connMaster.tableOperations().create(masterTable2);
        String masterTableId2 = connMaster.tableOperations().tableIdMap().get(masterTable2);
        Assert.assertNotNull(masterTableId2);
        connPeer.tableOperations().create(peerTable1);
        String peerTableId1 = connPeer.tableOperations().tableIdMap().get(peerTable1);
        Assert.assertNotNull(peerTableId1);
        connPeer.tableOperations().create(peerTable2);
        String peerTableId2 = connPeer.tableOperations().tableIdMap().get(peerTable2);
        Assert.assertNotNull(peerTableId2);
        // Give our replication user the ability to write to the tables
        connPeer.securityOperations().grantTablePermission(peerUserName, peerTable1, TablePermission.WRITE);
        connPeer.securityOperations().grantTablePermission(peerUserName, peerTable2, TablePermission.WRITE);
        // Replicate this table to the peerClusterName in a table with the peerTableId table id
        connMaster.tableOperations().setProperty(masterTable1, Property.TABLE_REPLICATION.getKey(), "true");
        connMaster.tableOperations().setProperty(masterTable1, Property.TABLE_REPLICATION_TARGET.getKey() + peerClusterName, peerTableId1);
        connMaster.tableOperations().setProperty(masterTable2, Property.TABLE_REPLICATION.getKey(), "true");
        connMaster.tableOperations().setProperty(masterTable2, Property.TABLE_REPLICATION_TARGET.getKey() + peerClusterName, peerTableId2);
        // Write some data to table1
        BatchWriter bw = connMaster.createBatchWriter(masterTable1, new BatchWriterConfig());
        for (int rows = 0; rows < 2500; rows++) {
            Mutation m = new Mutation(masterTable1 + rows);
            for (int cols = 0; cols < 100; cols++) {
                String value = Integer.toString(cols);
                m.put(value, "", value);
            }
            bw.addMutation(m);
        }
        bw.close();
        // Write some data to table2
        bw = connMaster.createBatchWriter(masterTable2, new BatchWriterConfig());
        for (int rows = 0; rows < 2500; rows++) {
            Mutation m = new Mutation(masterTable2 + rows);
            for (int cols = 0; cols < 100; cols++) {
                String value = Integer.toString(cols);
                m.put(value, "", value);
            }
            bw.addMutation(m);
        }
        bw.close();
        log.info("Wrote all data to master cluster");
        for (ProcessReference proc : cluster.getProcesses().get(ServerType.TABLET_SERVER)) {
            cluster.killProcess(ServerType.TABLET_SERVER, proc);
        }
        cluster.exec(TabletServer.class);
        while (!ReplicationTable.isOnline(connMaster)) {
            log.info("Replication table still offline, waiting");
            Thread.sleep(5000);
        }
        // Wait until we fully replicated something
        boolean fullyReplicated = false;
        for (int i = 0; i < 10 && !fullyReplicated; i++) {
            sleepUninterruptibly(2, TimeUnit.SECONDS);
            try (Scanner s = ReplicationTable.getScanner(connMaster)) {
                WorkSection.limit(s);
                for (Entry<Key, Value> entry : s) {
                    Status status = Status.parseFrom(entry.getValue().get());
                    if (StatusUtil.isFullyReplicated(status)) {
                        fullyReplicated |= true;
                    }
                }
            }
        }
        Assert.assertNotEquals(0, fullyReplicated);
        // We have to wait for the master to assign the replication work, a local tserver to process it, and then the remote tserver to replay it
        // Be cautious in how quickly we assert that the data is present on the peer
        long countTable = 0l;
        for (int i = 0; i < 10; i++) {
            for (Entry<Key, Value> entry : connPeer.createScanner(peerTable1, Authorizations.EMPTY)) {
                countTable++;
                Assert.assertTrue("Found unexpected key-value" + entry.getKey().toStringNoTruncate() + " " + entry.getValue(), entry.getKey().getRow().toString().startsWith(masterTable1));
            }
            log.info("Found {} records in {}", countTable, peerTable1);
            if (0l == countTable) {
                Thread.sleep(5000);
            } else {
                break;
            }
        }
        Assert.assertTrue("Found no records in " + peerTable1 + " in the peer cluster", countTable > 0);
        // Be cautious in how quickly we assert that the data is present on the peer
        for (int i = 0; i < 10; i++) {
            countTable = 0l;
            for (Entry<Key, Value> entry : connPeer.createScanner(peerTable2, Authorizations.EMPTY)) {
                countTable++;
                Assert.assertTrue("Found unexpected key-value" + entry.getKey().toStringNoTruncate() + " " + entry.getValue(), entry.getKey().getRow().toString().startsWith(masterTable2));
            }
            log.info("Found {} records in {}", countTable, peerTable2);
            if (0l == countTable) {
                Thread.sleep(5000);
            } else {
                break;
            }
        }
        Assert.assertTrue("Found no records in " + peerTable2 + " in the peer cluster", countTable > 0);
    } finally {
        peer1Cluster.stop();
    }
}

Example 3 with Status

use of org.apache.accumulo.server.replication.proto.Replication.Status in project accumulo by apache.

the class ReplicationIT method singleTableWithSingleTarget.

@Test
public void singleTableWithSingleTarget() throws Exception {
    // We want to kill the GC so it doesn't come along and close Status records and mess up the comparisons
    // against expected Status messages.
    getCluster().getClusterControl().stop(ServerType.GARBAGE_COLLECTOR);
    Connector conn = getConnector();
    String table1 = "table1";
    // replication shouldn't be online when we begin
    Assert.assertFalse(ReplicationTable.isOnline(conn));
    // Create a table
    conn.tableOperations().create(table1);
    int attempts = 10;
    // Might think the table doesn't yet exist, retry
    while (attempts > 0) {
        try {
            // Enable replication on table1
            conn.tableOperations().setProperty(table1, Property.TABLE_REPLICATION.getKey(), "true");
            // Replicate table1 to cluster1 in the table with id of '4'
            conn.tableOperations().setProperty(table1, Property.TABLE_REPLICATION_TARGET.getKey() + "cluster1", "4");
            // Sleep for 100 seconds before saying something is replicated
            conn.instanceOperations().setProperty(Property.REPLICATION_PEERS.getKey() + "cluster1", ReplicaSystemFactory.getPeerConfigurationValue(MockReplicaSystem.class, "100000"));
            break;
        } catch (Exception e) {
            attempts--;
            if (attempts <= 0) {
                throw e;
            }
            sleepUninterruptibly(2, TimeUnit.SECONDS);
        }
    }
    // Write some data to table1
    writeSomeData(conn, table1, 2000, 50);
    // Make sure the replication table is online at this point
    while (!ReplicationTable.isOnline(conn)) {
        sleepUninterruptibly(MILLIS_BETWEEN_REPLICATION_TABLE_ONLINE_CHECKS, TimeUnit.MILLISECONDS);
    }
    Assert.assertTrue("Replication table was never created", ReplicationTable.isOnline(conn));
    // ACCUMULO-2743 The Observer in the tserver has to be made aware of the change to get the combiner (made by the master)
    for (int i = 0; i < 10 && !conn.tableOperations().listIterators(ReplicationTable.NAME).keySet().contains(ReplicationTable.COMBINER_NAME); i++) {
        sleepUninterruptibly(2, TimeUnit.SECONDS);
    }
    Assert.assertTrue("Combiner was never set on replication table", conn.tableOperations().listIterators(ReplicationTable.NAME).keySet().contains(ReplicationTable.COMBINER_NAME));
    // Trigger the minor compaction, waiting for it to finish.
    // This should write the entry to metadata that the file has data
    conn.tableOperations().flush(table1, null, null, true);
    // Make sure that we have one status element, should be a new file
    try (Scanner s = ReplicationTable.getScanner(conn)) {
        StatusSection.limit(s);
        Entry<Key, Value> entry = null;
        Status expectedStatus = StatusUtil.openWithUnknownLength();
        attempts = 10;
        // This record will move from new to new with infinite length because of the minc (flush)
        while (null == entry && attempts > 0) {
            try {
                entry = Iterables.getOnlyElement(s);
                Status actual = Status.parseFrom(entry.getValue().get());
                if (actual.getInfiniteEnd() != expectedStatus.getInfiniteEnd()) {
                    entry = null;
                    // the master process didn't yet fire and write the new mutation, wait for it to do
                    // so and try to read it again
                    Thread.sleep(1000);
                }
            } catch (NoSuchElementException e) {
                entry = null;
                Thread.sleep(500);
            } catch (IllegalArgumentException e) {
                // saw this contain 2 elements once
                try (Scanner s2 = ReplicationTable.getScanner(conn)) {
                    StatusSection.limit(s2);
                    for (Entry<Key, Value> content : s2) {
                        log.info("{} => {}", content.getKey().toStringNoTruncate(), content.getValue());
                    }
                    throw e;
                }
            } finally {
                attempts--;
            }
        }
        Assert.assertNotNull("Could not find expected entry in replication table", entry);
        Status actual = Status.parseFrom(entry.getValue().get());
        Assert.assertTrue("Expected to find a replication entry that is open with infinite length: " + ProtobufUtil.toString(actual), !actual.getClosed() && actual.getInfiniteEnd());
        // Try a couple of times to watch for the work record to be created
        boolean notFound = true;
        for (int i = 0; i < 10 && notFound; i++) {
            try (Scanner s2 = ReplicationTable.getScanner(conn)) {
                WorkSection.limit(s2);
                int elementsFound = Iterables.size(s2);
                if (0 < elementsFound) {
                    Assert.assertEquals(1, elementsFound);
                    notFound = false;
                }
                Thread.sleep(500);
            }
        }
        // If we didn't find the work record, print the contents of the table
        if (notFound) {
            try (Scanner s2 = ReplicationTable.getScanner(conn)) {
                for (Entry<Key, Value> content : s2) {
                    log.info("{} => {}", content.getKey().toStringNoTruncate(), content.getValue());
                }
                Assert.assertFalse("Did not find the work entry for the status entry", notFound);
            }
        }
        // Write some more data so that we over-run the single WAL
        writeSomeData(conn, table1, 3000, 50);
        log.info("Issued compaction for table");
        conn.tableOperations().compact(table1, null, null, true, true);
        log.info("Compaction completed");
        // Master is creating entries in the replication table from the metadata table every second.
        // Compaction should trigger the record to be written to metadata. Wait a bit to ensure
        // that the master has time to work.
        Thread.sleep(5000);
        try (Scanner s2 = ReplicationTable.getScanner(conn)) {
            StatusSection.limit(s2);
            int numRecords = 0;
            for (Entry<Key, Value> e : s2) {
                numRecords++;
                log.info("Found status record {}\t{}", e.getKey().toStringNoTruncate(), ProtobufUtil.toString(Status.parseFrom(e.getValue().get())));
            }
            Assert.assertEquals(2, numRecords);
        }
        // We should eventually get 2 work records recorded, need to account for a potential delay though
        // might see: status1 -> work1 -> status2 -> (our scans) -> work2
        notFound = true;
        for (int i = 0; i < 10 && notFound; i++) {
            try (Scanner s2 = ReplicationTable.getScanner(conn)) {
                WorkSection.limit(s2);
                int elementsFound = Iterables.size(s2);
                if (2 == elementsFound) {
                    notFound = false;
                }
                Thread.sleep(500);
            }
        }
        // If we didn't find the work record, print the contents of the table
        if (notFound) {
            try (Scanner s2 = ReplicationTable.getScanner(conn)) {
                for (Entry<Key, Value> content : s2) {
                    log.info("{} => {}", content.getKey().toStringNoTruncate(), content.getValue());
                }
                Assert.assertFalse("Did not find the work entries for the status entries", notFound);
            }
        }
    }
}

Example 4 with Status

use of org.apache.accumulo.server.replication.proto.Replication.Status in project accumulo by apache.

the class ReplicationIT method filesClosedAfterUnused.

@Test
public void filesClosedAfterUnused() throws Exception {
    Connector conn = getConnector();
    String table = "table";
    conn.tableOperations().create(table);
    Table.ID tableId = Table.ID.of(conn.tableOperations().tableIdMap().get(table));
    Assert.assertNotNull(tableId);
    conn.tableOperations().setProperty(table, Property.TABLE_REPLICATION.getKey(), "true");
    conn.tableOperations().setProperty(table, Property.TABLE_REPLICATION_TARGET.getKey() + "cluster1", "1");
    // just sleep
    conn.instanceOperations().setProperty(Property.REPLICATION_PEERS.getKey() + "cluster1", ReplicaSystemFactory.getPeerConfigurationValue(MockReplicaSystem.class, "50000"));
    // Write a mutation to make a log file
    BatchWriter bw = conn.createBatchWriter(table, new BatchWriterConfig());
    Mutation m = new Mutation("one");
    m.put("", "", "");
    bw.addMutation(m);
    bw.close();
    // Write another to make sure the logger rolls itself?
    bw = conn.createBatchWriter(table, new BatchWriterConfig());
    m = new Mutation("three");
    m.put("", "", "");
    bw.addMutation(m);
    bw.close();
    try (Scanner s = conn.createScanner(MetadataTable.NAME, Authorizations.EMPTY)) {
        s.fetchColumnFamily(TabletsSection.LogColumnFamily.NAME);
        s.setRange(TabletsSection.getRange(tableId));
        Set<String> wals = new HashSet<>();
        for (Entry<Key, Value> entry : s) {
            LogEntry logEntry = LogEntry.fromKeyValue(entry.getKey(), entry.getValue());
            wals.add(new Path(logEntry.filename).toString());
        }
        log.warn("Found wals {}", wals);
        bw = conn.createBatchWriter(table, new BatchWriterConfig());
        m = new Mutation("three");
        byte[] bytes = new byte[1024 * 1024];
        m.put("1".getBytes(), new byte[0], bytes);
        m.put("2".getBytes(), new byte[0], bytes);
        m.put("3".getBytes(), new byte[0], bytes);
        m.put("4".getBytes(), new byte[0], bytes);
        m.put("5".getBytes(), new byte[0], bytes);
        bw.addMutation(m);
        bw.close();
        conn.tableOperations().flush(table, null, null, true);
        while (!ReplicationTable.isOnline(conn)) {
            sleepUninterruptibly(MILLIS_BETWEEN_REPLICATION_TABLE_ONLINE_CHECKS, TimeUnit.MILLISECONDS);
        }
        for (int i = 0; i < 10; i++) {
            try (Scanner s2 = conn.createScanner(MetadataTable.NAME, Authorizations.EMPTY)) {
                s2.fetchColumnFamily(LogColumnFamily.NAME);
                s2.setRange(TabletsSection.getRange(tableId));
                for (Entry<Key, Value> entry : s2) {
                    log.info("{}={}", entry.getKey().toStringNoTruncate(), entry.getValue());
                }
            }
            try (Scanner s3 = ReplicationTable.getScanner(conn)) {
                StatusSection.limit(s3);
                Text buff = new Text();
                boolean allReferencedLogsClosed = true;
                int recordsFound = 0;
                for (Entry<Key, Value> e : s3) {
                    recordsFound++;
                    allReferencedLogsClosed = true;
                    StatusSection.getFile(e.getKey(), buff);
                    String file = buff.toString();
                    if (wals.contains(file)) {
                        Status stat = Status.parseFrom(e.getValue().get());
                        if (!stat.getClosed()) {
                            log.info("{} wasn't closed", file);
                            allReferencedLogsClosed = false;
                        }
                    }
                }
                if (recordsFound > 0 && allReferencedLogsClosed) {
                    return;
                }
                Thread.sleep(2000);
            } catch (RuntimeException e) {
                Throwable cause = e.getCause();
                if (cause instanceof AccumuloSecurityException) {
                    AccumuloSecurityException ase = (AccumuloSecurityException) cause;
                    switch(ase.getSecurityErrorCode()) {
                        case PERMISSION_DENIED:
                            // We tried to read the replication table before the GRANT went through
                            Thread.sleep(2000);
                            break;
                        default:
                            throw e;
                    }
                }
            }
        }
        Assert.fail("We had a file that was referenced but didn't get closed");
    }
}

Example 5 with Status

use of org.apache.accumulo.server.replication.proto.Replication.Status in project accumulo by apache.

the class ReplicationOperationsImplIT method waitsUntilEntriesAreReplicated.

@Test
public void waitsUntilEntriesAreReplicated() throws Exception {
    conn.tableOperations().create("foo");
    Table.ID tableId = Table.ID.of(conn.tableOperations().tableIdMap().get("foo"));
    String file1 = "/accumulo/wals/tserver+port/" + UUID.randomUUID(), file2 = "/accumulo/wals/tserver+port/" + UUID.randomUUID();
    Status stat = Status.newBuilder().setBegin(0).setEnd(10000).setInfiniteEnd(false).setClosed(false).build();
    BatchWriter bw = ReplicationTable.getBatchWriter(conn);
    Mutation m = new Mutation(file1);
    StatusSection.add(m, tableId, ProtobufUtil.toValue(stat));
    bw.addMutation(m);
    m = new Mutation(file2);
    StatusSection.add(m, tableId, ProtobufUtil.toValue(stat));
    bw.addMutation(m);
    bw.close();
    bw = conn.createBatchWriter(MetadataTable.NAME, new BatchWriterConfig());
    m = new Mutation(ReplicationSection.getRowPrefix() + file1);
    m.put(ReplicationSection.COLF, new Text(tableId.getUtf8()), ProtobufUtil.toValue(stat));
    bw.addMutation(m);
    m = new Mutation(ReplicationSection.getRowPrefix() + file2);
    m.put(ReplicationSection.COLF, new Text(tableId.getUtf8()), ProtobufUtil.toValue(stat));
    bw.close();
    final AtomicBoolean done = new AtomicBoolean(false);
    final AtomicBoolean exception = new AtomicBoolean(false);
    final ReplicationOperationsImpl roi = getReplicationOperations();
    Thread t = new Thread(new Runnable() {

        @Override
        public void run() {
            try {
                roi.drain("foo");
            } catch (Exception e) {
                log.error("Got error", e);
                exception.set(true);
            }
            done.set(true);
        }
    });
    t.start();
    // With the records, we shouldn't be drained
    Assert.assertFalse(done.get());
    bw = conn.createBatchWriter(MetadataTable.NAME, new BatchWriterConfig());
    m = new Mutation(ReplicationSection.getRowPrefix() + file1);
    m.putDelete(ReplicationSection.COLF, new Text(tableId.getUtf8()));
    bw.addMutation(m);
    bw.flush();
    Assert.assertFalse(done.get());
    m = new Mutation(ReplicationSection.getRowPrefix() + file2);
    m.putDelete(ReplicationSection.COLF, new Text(tableId.getUtf8()));
    bw.addMutation(m);
    bw.flush();
    bw.close();
    // Removing metadata entries doesn't change anything
    Assert.assertFalse(done.get());
    // Remove the replication entries too
    bw = ReplicationTable.getBatchWriter(conn);
    m = new Mutation(file1);
    m.putDelete(StatusSection.NAME, new Text(tableId.getUtf8()));
    bw.addMutation(m);
    bw.flush();
    Assert.assertFalse(done.get());
    m = new Mutation(file2);
    m.putDelete(StatusSection.NAME, new Text(tableId.getUtf8()));
    bw.addMutation(m);
    bw.flush();
    try {
        t.join(5000);
    } catch (InterruptedException e) {
        Assert.fail("ReplicationOperations.drain did not complete");
    }
    // After both metadata and replication
    Assert.assertTrue("Drain never finished", done.get());
    Assert.assertFalse("Saw unexpectetd exception", exception.get());
}