use of com.couchbase.connector.dcp.CheckpointDao in project couchbase-elasticsearch-connector by couchbase.
the class CheckpointBackup method backup.
public static void backup(ConnectorConfig config, File outputFile) throws IOException {
final Bucket bucket = CouchbaseHelper.openMetadataBucket(config.couchbase(), config.trustStore());
final ResolvedBucketConfig bucketConfig = getBucketConfig(config.couchbase(), bucket);
// don't care bucketConfig.uuid();
final String bucketUuid = "";
final CheckpointDao checkpointDao = new CouchbaseCheckpointDao(getMetadataCollection(bucket, config.couchbase()), config.group().name());
final int numVbuckets = bucketConfig.numberOfPartitions();
final Set<Integer> vbuckets = IntStream.range(0, numVbuckets).boxed().collect(toSet());
final Map<Integer, Checkpoint> checkpoints = checkpointDao.load(bucketUuid, vbuckets);
final Map<String, Object> output = new LinkedHashMap<>();
output.put("formatVersion", 1);
output.put("bucketUuid", bucketConfig.uuid());
output.put("vbuckets", checkpoints);
atomicWrite(outputFile, tempFile -> {
try (FileOutputStream out = new FileOutputStream(tempFile)) {
new ObjectMapper().writeValue(out, output);
}
});
System.out.println("Wrote checkpoint for connector '" + config.group().name() + "' to file " + outputFile.getAbsolutePath());
}
use of com.couchbase.connector.dcp.CheckpointDao in project couchbase-elasticsearch-connector by couchbase.
the class CheckpointRestore method restore.
public static void restore(ConnectorConfig config, File inputFile) throws IOException {
final Bucket bucket = CouchbaseHelper.openMetadataBucket(config.couchbase(), config.trustStore());
final ResolvedBucketConfig bucketConfig = getBucketConfig(config.couchbase(), bucket);
final CheckpointDao checkpointDao = new CouchbaseCheckpointDao(getMetadataCollection(bucket, config.couchbase()), config.group().name());
final ObjectMapper mapper = new ObjectMapper();
try (InputStream is = new FileInputStream(inputFile)) {
final JsonNode json = mapper.readTree(is);
final int formatVersion = json.path("formatVersion").intValue();
final String bucketUuid = json.path("bucketUuid").asText();
if (formatVersion != 1) {
throw new IllegalArgumentException("Unrecognized checkpoint format version: " + formatVersion);
}
if (!bucketUuid.equals(bucketConfig.uuid())) {
throw new IllegalArgumentException("Bucket UUID mismatch; checkpoint is from a bucket with UUID " + bucketUuid + " but this bucket has UUID " + bucketConfig.uuid());
}
final Map<Integer, Checkpoint> checkpoints = mapper.convertValue(json.get("vbuckets"), new TypeReference<Map<Integer, Checkpoint>>() {
});
if (checkpoints.size() != bucketConfig.numberOfPartitions()) {
throw new IllegalArgumentException("Bucket has " + bucketConfig.numberOfPartitions() + " vbuckets but the checkpoint file has " + checkpoints.size() + " -- is it from a different operating system (for example, macOS vs Linux)?");
}
checkpointDao.save("", checkpoints);
}
System.out.println("Restored checkpoint for connector '" + config.group().name() + "' from file " + inputFile);
}
use of com.couchbase.connector.dcp.CheckpointDao in project couchbase-elasticsearch-connector by couchbase.
the class ElasticsearchConnector method run.
public static void run(ConnectorConfig config, PanicButton panicButton, Duration startupQuietPeriod) throws Throwable {
final Throwable fatalError;
final Membership membership = config.group().staticMembership();
LOGGER.info("Read configuration: {}", redactSystem(config));
final ScheduledExecutorService checkpointExecutor = Executors.newSingleThreadScheduledExecutor();
try (Slf4jReporter metricReporter = newSlf4jReporter(config.metrics().logInterval());
HttpServer httpServer = new HttpServer(config.metrics().httpPort(), membership);
RestHighLevelClient esClient = newElasticsearchClient(config.elasticsearch(), config.trustStore())) {
DocumentLifecycle.setLogLevel(config.logging().logDocumentLifecycle() ? LogLevel.INFO : LogLevel.DEBUG);
LogRedaction.setRedactionLevel(config.logging().redactionLevel());
DcpHelper.setRedactionLevel(config.logging().redactionLevel());
final ClusterEnvironment env = CouchbaseHelper.environmentBuilder(config.couchbase(), config.trustStore()).build();
final Cluster cluster = CouchbaseHelper.createCluster(config.couchbase(), env);
final Version elasticsearchVersion = waitForElasticsearchAndRequireVersion(esClient, new Version(2, 0, 0), new Version(5, 6, 16));
LOGGER.info("Elasticsearch version {}", elasticsearchVersion);
validateConfig(elasticsearchVersion, config.elasticsearch());
// Wait for couchbase server to come online, then open the bucket.
final Bucket bucket = CouchbaseHelper.waitForBucket(cluster, config.couchbase().bucket());
final Set<SeedNode> kvNodes = CouchbaseHelper.getKvNodes(config.couchbase(), bucket);
final boolean storeMetadataInSourceBucket = config.couchbase().metadataBucket().equals(config.couchbase().bucket());
final Bucket metadataBucket = storeMetadataInSourceBucket ? bucket : CouchbaseHelper.waitForBucket(cluster, config.couchbase().metadataBucket());
final Collection metadataCollection = CouchbaseHelper.getMetadataCollection(metadataBucket, config.couchbase());
final CheckpointDao checkpointDao = new CouchbaseCheckpointDao(metadataCollection, config.group().name());
// todo get this from dcp client
final String bucketUuid = "";
final CheckpointService checkpointService = new CheckpointService(bucketUuid, checkpointDao);
final RequestFactory requestFactory = new RequestFactory(config.elasticsearch().types(), config.elasticsearch().docStructure(), config.elasticsearch().rejectLog());
final ElasticsearchWorkerGroup workers = new ElasticsearchWorkerGroup(esClient, checkpointService, requestFactory, ErrorListener.NOOP, config.elasticsearch().bulkRequest());
Metrics.gauge("write.queue", "Document events currently buffered in memory.", workers, ElasticsearchWorkerGroup::getQueueSize);
// High value indicates the connector has stalled
Metrics.gauge("es.wait.ms", null, workers, ElasticsearchWorkerGroup::getCurrentRequestMillis);
// Same as "es.wait.ms" but normalized to seconds for Prometheus
Metrics.gauge("es.wait.seconds", "Duration of in-flight Elasticsearch bulk request (including any retries). Long duration may indicate connector has stalled.", workers, value -> value.getCurrentRequestMillis() / (double) SECONDS.toMillis(1));
final Client dcpClient = DcpHelper.newClient(config.group().name(), config.couchbase(), kvNodes, config.trustStore());
initEventListener(dcpClient, panicButton, workers::submit);
final Thread saveCheckpoints = new Thread(checkpointService::save, "save-checkpoints");
try {
try {
dcpClient.connect().block(Duration.ofMillis(config.couchbase().dcp().connectTimeout().millis()));
} catch (Throwable t) {
panicButton.panic("Failed to establish initial DCP connection within " + config.couchbase().dcp().connectTimeout(), t);
}
final int numPartitions = dcpClient.numPartitions();
LOGGER.info("Bucket has {} partitions. Membership = {}", numPartitions, membership);
final Set<Integer> partitions = membership.getPartitions(numPartitions);
if (partitions.isEmpty()) {
// need to do this check, because if we started streaming with an empty list, the DCP client would open streams for *all* partitions
throw new IllegalArgumentException("There are more workers than Couchbase vbuckets; this worker doesn't have any work to do.");
}
checkpointService.init(numPartitions, () -> DcpHelper.getCurrentSeqnosAsMap(dcpClient, partitions, Duration.ofSeconds(5)));
dcpClient.initializeState(StreamFrom.BEGINNING, StreamTo.INFINITY).block();
initSessionState(dcpClient, checkpointService, partitions);
// configuration problems.
if (!startupQuietPeriod.isZero()) {
LOGGER.info("Entering startup quiet period; sleeping for {} so peers can terminate in case of unsafe scaling.", startupQuietPeriod);
MILLISECONDS.sleep(startupQuietPeriod.toMillis());
LOGGER.info("Startup quiet period complete.");
}
checkpointExecutor.scheduleWithFixedDelay(checkpointService::save, 10, 10, SECONDS);
RuntimeHelper.addShutdownHook(saveCheckpoints);
// Unless shutdown is due to panic...
panicButton.addPrePanicHook(() -> RuntimeHelper.removeShutdownHook(saveCheckpoints));
try {
LOGGER.debug("Opening DCP streams for partitions: {}", partitions);
dcpClient.startStreaming(partitions).block();
} catch (RuntimeException e) {
ThrowableHelper.propagateCauseIfPossible(e, InterruptedException.class);
throw e;
}
// Start HTTP server *after* other setup is complete, so the metrics endpoint
// can be used as a "successful startup" probe.
httpServer.start();
if (config.metrics().httpPort() >= 0) {
LOGGER.info("Prometheus metrics available at http://localhost:{}/metrics/prometheus", httpServer.getBoundPort());
LOGGER.info("Dropwizard metrics available at http://localhost:{}/metrics/dropwizard?pretty", httpServer.getBoundPort());
} else {
LOGGER.info("Metrics HTTP server is disabled. Edit the [metrics] 'httpPort' config property to enable.");
}
LOGGER.info("Elasticsearch connector startup complete.");
fatalError = workers.awaitFatalError();
LOGGER.error("Terminating due to fatal error from worker", fatalError);
} catch (InterruptedException shutdownRequest) {
LOGGER.info("Graceful shutdown requested. Saving checkpoints and cleaning up.");
checkpointService.save();
throw shutdownRequest;
} catch (Throwable t) {
LOGGER.error("Terminating due to fatal error during setup", t);
throw t;
} finally {
// If we get here it means there was a fatal exception, or the connector is running in distributed
// or test mode and a graceful shutdown was requested. Don't need the shutdown hook for any of those cases.
RuntimeHelper.removeShutdownHook(saveCheckpoints);
checkpointExecutor.shutdown();
metricReporter.stop();
dcpClient.disconnect().block();
// to avoid buffer leak, must close *after* dcp client stops feeding it events
workers.close();
checkpointExecutor.awaitTermination(10, SECONDS);
cluster.disconnect();
// can't reuse, because connector config might have different SSL settings next time
env.shutdown();
}
}
// give stdout a chance to quiet down so the stack trace on stderr isn't interleaved with stdout.
MILLISECONDS.sleep(500);
throw fatalError;
}
use of com.couchbase.connector.dcp.CheckpointDao in project couchbase-elasticsearch-connector by couchbase.
the class CheckpointClear method run.
private static void run(ConnectorConfig config, boolean catchUp) throws IOException {
final ClusterEnvironment env = environmentBuilder(config.couchbase(), config.trustStore()).build();
final Cluster cluster = createCluster(config.couchbase(), env);
try {
final Bucket metadataBucket = CouchbaseHelper.waitForBucket(cluster, config.couchbase().metadataBucket());
final Collection metadataCollection = CouchbaseHelper.getMetadataCollection(metadataBucket, config.couchbase());
final ResolvedBucketConfig bucketConfig = getBucketConfig(config.couchbase(), metadataBucket);
final Set<SeedNode> kvNodes = getKvNodes(config.couchbase(), metadataBucket);
final CheckpointDao checkpointDao = new CouchbaseCheckpointDao(metadataCollection, config.group().name());
if (catchUp) {
setCheckpointToNow(config, kvNodes, checkpointDao);
System.out.println("Set checkpoint for connector '" + config.group().name() + "' to match current state of Couchbase bucket.");
} else {
final int numVbuckets = bucketConfig.numberOfPartitions();
final Set<Integer> vbuckets = IntStream.range(0, numVbuckets).boxed().collect(toSet());
checkpointDao.clear(bucketConfig.uuid(), vbuckets);
System.out.println("Cleared checkpoint for connector '" + config.group().name() + "'.");
}
} finally {
cluster.disconnect();
env.shutdown();
}
}
Aggregations