Examples with SnapshotInfo org.apache.ratis.statemachine.SnapshotInfo used on opensource projects

Example 16 with SnapshotInfo

use of org.apache.ratis.statemachine.SnapshotInfo in project incubator-ratis by apache.

the class LogAppender method shouldInstallSnapshot.

/**
 * Should this {@link LogAppender} send a snapshot to the follower?
 *
 * @return the snapshot if it should install a snapshot; otherwise, return null.
 */
default SnapshotInfo shouldInstallSnapshot() {
    // we should install snapshot if the follower needs to catch up and:
    // 1. there is no local log entry but there is snapshot
    // 2. or the follower's next index is smaller than the log start index
    // 3. or the follower is bootstrapping and has not installed any snapshot yet
    final FollowerInfo follower = getFollower();
    final boolean isFollowerBootstrapping = getLeaderState().isFollowerBootstrapping(follower);
    final SnapshotInfo snapshot = getServer().getStateMachine().getLatestSnapshot();
    if (isFollowerBootstrapping && !follower.hasAttemptedToInstallSnapshot()) {
        if (snapshot == null) {
            // Leader cannot send null snapshot to follower. Hence, acknowledge InstallSnapshot attempt (even though it
            // was not attempted) so that follower can come out of staging state after appending log entries.
            follower.setAttemptedToInstallSnapshot();
        } else {
            return snapshot;
        }
    }
    final long followerNextIndex = getFollower().getNextIndex();
    if (followerNextIndex < getRaftLog().getNextIndex()) {
        final long logStartIndex = getRaftLog().getStartIndex();
        if (followerNextIndex < logStartIndex || (logStartIndex == RaftLog.INVALID_LOG_INDEX && snapshot != null)) {
            return snapshot;
        }
    }
    return null;
}

Example 17 with SnapshotInfo

use of org.apache.ratis.statemachine.SnapshotInfo in project incubator-ratis by apache.

the class SnapshotManager method installSnapshot.

@SuppressFBWarnings("RV_RETURN_VALUE_IGNORED_BAD_PRACTICE")
public void installSnapshot(StateMachine stateMachine, InstallSnapshotRequestProto request) throws IOException {
    final InstallSnapshotRequestProto.SnapshotChunkProto snapshotChunkRequest = request.getSnapshotChunk();
    final long lastIncludedIndex = snapshotChunkRequest.getTermIndex().getIndex();
    final RaftStorageDirectory dir = storage.getStorageDir();
    // create a unique temporary directory
    final File tmpDir = new File(dir.getTmpDir(), UUID.randomUUID().toString());
    FileUtils.createDirectories(tmpDir);
    tmpDir.deleteOnExit();
    LOG.info("Installing snapshot:{}, to tmp dir:{}", request, tmpDir);
    for (FileChunkProto chunk : snapshotChunkRequest.getFileChunksList()) {
        SnapshotInfo pi = stateMachine.getLatestSnapshot();
        if (pi != null && pi.getTermIndex().getIndex() >= lastIncludedIndex) {
            throw new IOException("There exists snapshot file " + pi.getFiles() + " in " + selfId + " with endIndex >= lastIncludedIndex " + lastIncludedIndex);
        }
        // this is relative to the root dir
        String fileName = chunk.getFilename();
        // TODO: assumes flat layout inside SM dir
        File tmpSnapshotFile = new File(tmpDir, new File(dir.getRoot(), fileName).getName());
        FileOutputStream out = null;
        try {
            // same last index.
            if (chunk.getOffset() == 0) {
                if (tmpSnapshotFile.exists()) {
                    FileUtils.deleteFully(tmpSnapshotFile);
                }
                // create the temp snapshot file and put padding inside
                out = new FileOutputStream(tmpSnapshotFile);
            } else {
                Preconditions.assertTrue(tmpSnapshotFile.exists());
                out = new FileOutputStream(tmpSnapshotFile, true);
                FileChannel fc = out.getChannel();
                fc.position(chunk.getOffset());
            }
            // write data to the file
            out.write(chunk.getData().toByteArray());
        } finally {
            IOUtils.cleanup(null, out);
        }
        // the md5 digest and create the md5 meta-file.
        if (chunk.getDone()) {
            final MD5Hash expectedDigest = new MD5Hash(chunk.getFileDigest().toByteArray());
            // calculate the checksum of the snapshot file and compare it with the
            // file digest in the request
            MD5Hash digest = MD5FileUtil.computeMd5ForFile(tmpSnapshotFile);
            if (!digest.equals(expectedDigest)) {
                LOG.warn("The snapshot md5 digest {} does not match expected {}", digest, expectedDigest);
                // rename the temp snapshot file to .corrupt
                FileUtils.renameFileToCorrupt(tmpSnapshotFile);
                throw new CorruptedFileException(tmpSnapshotFile, "MD5 mismatch for snapshot-" + lastIncludedIndex + " installation");
            } else {
                MD5FileUtil.saveMD5File(tmpSnapshotFile, digest);
            }
        }
    }
    if (snapshotChunkRequest.getDone()) {
        LOG.info("Install snapshot is done, renaming tnp dir:{} to:{}", tmpDir, dir.getStateMachineDir());
        dir.getStateMachineDir().delete();
        tmpDir.renameTo(dir.getStateMachineDir());
    }
}

Example 18 with SnapshotInfo

use of org.apache.ratis.statemachine.SnapshotInfo in project incubator-ratis by apache.

the class InstallSnapshotNotificationTests method testAddNewFollowers.

private void testAddNewFollowers(CLUSTER cluster) throws Exception {
    leaderSnapshotInfoRef.set(null);
    final List<LogSegmentPath> logs;
    int i = 0;
    try {
        RaftTestUtil.waitForLeader(cluster);
        final RaftPeerId leaderId = cluster.getLeader().getId();
        try (final RaftClient client = cluster.createClient(leaderId)) {
            for (; i < SNAPSHOT_TRIGGER_THRESHOLD * 2 - 1; i++) {
                RaftClientReply reply = client.io().send(new RaftTestUtil.SimpleMessage("m" + i));
                Assert.assertTrue(reply.isSuccess());
            }
        }
        // wait for the snapshot to be done
        final RaftServer.Division leader = cluster.getLeader();
        final long nextIndex = leader.getRaftLog().getNextIndex();
        LOG.info("nextIndex = {}", nextIndex);
        final List<File> snapshotFiles = RaftSnapshotBaseTest.getSnapshotFiles(cluster, nextIndex - SNAPSHOT_TRIGGER_THRESHOLD, nextIndex);
        JavaUtils.attemptRepeatedly(() -> {
            Assert.assertTrue(snapshotFiles.stream().anyMatch(RaftSnapshotBaseTest::exists));
            return null;
        }, 10, ONE_SECOND, "snapshotFile.exist", LOG);
        logs = LogSegmentPath.getLogSegmentPaths(leader.getRaftStorage());
    } finally {
        cluster.shutdown();
    }
    // delete the log segments from the leader
    LOG.info("Delete logs {}", logs);
    for (LogSegmentPath path : logs) {
        // the log may be already puged
        FileUtils.deleteFully(path.getPath());
    }
    // restart the peer
    LOG.info("Restarting the cluster");
    cluster.restart(false);
    try {
        RaftSnapshotBaseTest.assertLeaderContent(cluster);
        // generate some more traffic
        try (final RaftClient client = cluster.createClient(cluster.getLeader().getId())) {
            Assert.assertTrue(client.io().send(new RaftTestUtil.SimpleMessage("m" + i)).isSuccess());
        }
        final SnapshotInfo leaderSnapshotInfo = cluster.getLeader().getStateMachine().getLatestSnapshot();
        final boolean set = leaderSnapshotInfoRef.compareAndSet(null, leaderSnapshotInfo);
        Assert.assertTrue(set);
        // add two more peers
        final MiniRaftCluster.PeerChanges change = cluster.addNewPeers(2, true, true);
        // trigger setConfiguration
        cluster.setConfiguration(change.allPeersInNewConf);
        RaftServerTestUtil.waitAndCheckNewConf(cluster, change.allPeersInNewConf, 0, null);
        // leader snapshot.
        for (RaftServer.Division follower : cluster.getFollowers()) {
            Assert.assertEquals(leaderSnapshotInfo.getIndex(), RaftServerTestUtil.getLatestInstalledSnapshotIndex(follower));
        }
        // restart the peer and check if it can correctly handle conf change
        cluster.restartServer(cluster.getLeader().getId(), false);
        RaftSnapshotBaseTest.assertLeaderContent(cluster);
    } finally {
        cluster.shutdown();
    }
}

Example 19 with SnapshotInfo

use of org.apache.ratis.statemachine.SnapshotInfo in project incubator-ratis by apache.

the class InstallSnapshotNotificationTests method testInstallSnapshotNotificationCount.

private void testInstallSnapshotNotificationCount(CLUSTER cluster) throws Exception {
    leaderSnapshotInfoRef.set(null);
    numSnapshotRequests.set(0);
    int i = 0;
    try {
        RaftTestUtil.waitForLeader(cluster);
        final RaftPeerId leaderId = cluster.getLeader().getId();
        // Let a few heartbeats pass.
        ONE_SECOND.sleep();
        Assert.assertEquals(0, numSnapshotRequests.get());
        // Generate data.
        try (final RaftClient client = cluster.createClient(leaderId)) {
            for (; i < 10; i++) {
                RaftClientReply reply = client.io().send(new RaftTestUtil.SimpleMessage("m" + i));
                Assert.assertTrue(reply.isSuccess());
            }
        }
        // Wait until index has been updated
        RaftTestUtil.waitFor(() -> cluster.getLeader().getStateMachine().getLastAppliedTermIndex().getIndex() == 20, 300, 15000);
        // Take snapshot and check result.
        long snapshotIndex = cluster.getLeader().getStateMachine().takeSnapshot();
        Assert.assertEquals(20, snapshotIndex);
        final SnapshotInfo leaderSnapshotInfo = cluster.getLeader().getStateMachine().getLatestSnapshot();
        Assert.assertEquals(20, leaderSnapshotInfo.getIndex());
        final boolean set = leaderSnapshotInfoRef.compareAndSet(null, leaderSnapshotInfo);
        Assert.assertTrue(set);
        // Wait for the snapshot to be done.
        final RaftServer.Division leader = cluster.getLeader();
        final long nextIndex = leader.getRaftLog().getNextIndex();
        Assert.assertEquals(21, nextIndex);
        // End index is exclusive.
        final List<File> snapshotFiles = RaftSnapshotBaseTest.getSnapshotFiles(cluster, 0, nextIndex);
        JavaUtils.attemptRepeatedly(() -> {
            Assert.assertTrue(snapshotFiles.stream().anyMatch(RaftSnapshotBaseTest::exists));
            return null;
        }, 10, ONE_SECOND, "snapshotFile.exist", LOG);
        // Clear all log files and reset cached log start index.
        long snapshotInstallIndex = leader.getRaftLog().onSnapshotInstalled(leader.getRaftLog().getLastCommittedIndex()).get();
        Assert.assertEquals(20, snapshotInstallIndex);
        // Check that logs are gone.
        Assert.assertEquals(0, LogSegmentPath.getLogSegmentPaths(leader.getRaftStorage()).size());
        Assert.assertEquals(RaftLog.INVALID_LOG_INDEX, leader.getRaftLog().getStartIndex());
        // Allow some heartbeats to go through, then make sure none of them had
        // snapshot requests.
        ONE_SECOND.sleep();
        Assert.assertEquals(0, numSnapshotRequests.get());
        // Make sure leader and followers are still up to date.
        for (RaftServer.Division follower : cluster.getFollowers()) {
            Assert.assertEquals(leader.getRaftLog().getNextIndex(), follower.getRaftLog().getNextIndex());
        }
        // Add two more peers who will need snapshots from the leader.
        final MiniRaftCluster.PeerChanges change = cluster.addNewPeers(2, true, true);
        // trigger setConfiguration
        cluster.setConfiguration(change.allPeersInNewConf);
        RaftServerTestUtil.waitAndCheckNewConf(cluster, change.allPeersInNewConf, 0, null);
        // Generate more data.
        try (final RaftClient client = cluster.createClient(leader.getId())) {
            Assert.assertTrue(client.io().send(new RaftTestUtil.SimpleMessage("m" + i)).isSuccess());
        }
        // Make sure leader and followers are still up to date.
        for (RaftServer.Division follower : cluster.getFollowers()) {
            // Give follower slightly time to catch up
            RaftTestUtil.waitFor(() -> leader.getRaftLog().getNextIndex() == follower.getRaftLog().getNextIndex(), 300, 15000);
        }
        // Make sure each new peer got one snapshot notification.
        Assert.assertEquals(2, numSnapshotRequests.get());
    } finally {
        cluster.shutdown();
    }
}

Example 20 with SnapshotInfo

use of org.apache.ratis.statemachine.SnapshotInfo in project alluxio by Alluxio.

the class SnapshotReplicationManager method handleRequest.

/**
 * Handles snapshot requests.
 *
 * @param queryRequest the query request
 * @return the response message, or null if the request is not handled
 * @throws IOException if any error occurred while handling the request
 */
public Message handleRequest(JournalQueryRequest queryRequest) throws IOException {
    if (queryRequest.hasSnapshotInfoRequest()) {
        SnapshotInfo latestSnapshot = mStorage.getLatestSnapshot();
        if (latestSnapshot == null) {
            LOG.debug("No snapshot to send");
            return toMessage(GetSnapshotInfoResponse.getDefaultInstance());
        }
        JournalQueryResponse response = JournalQueryResponse.newBuilder().setSnapshotInfoResponse(GetSnapshotInfoResponse.newBuilder().setLatest(toSnapshotMetadata(latestSnapshot.getTermIndex()))).build();
        LOG.debug("Sent snapshot info response {}", response);
        return toMessage(response);
    }
    if (queryRequest.hasSnapshotRequest()) {
        LOG.debug("Start sending snapshot to leader");
        sendSnapshotToLeader();
        return Message.EMPTY;
    }
    return null;
}