// make sure oozie changes mentioned FALCON-389 are done on the clusters. Otherwise the test
// will fail.
// Noticed with hive 0.13 we need the following issues resolved to work HIVE-6848 and
// HIVE-6868. Also oozie share libs need to have hive jars that have these jira's resolved and
// the maven depenendcy you are using to run the tests has to have hcat that has these fixed.
// This test can fail randomly because of https://issues.apache.org/jira/browse/FALCON-401
@Test(dataProvider = "generateSeparators")
public void oneSourceTwoTarget(String separator) throws Exception {
String tcName = "HCatReplication_oneSourceTwoTarget";
if (separator.equals("-")) {
tcName += "_hyphen";
} else {
tcName += "_slash";
}
String tblName = tcName;
String testHdfsDir = baseTestHDFSDir + "/" + tcName;
HadoopUtil.recreateDir(serverFS, testHdfsDir);
final String startDate = "2010-01-01T20:00Z";
final String endDate = "2099-01-01T00:00Z";
final String dataEndDate = "2010-01-01T21:00Z";
final String tableUriPartitionFragment =
StringUtils.join(new String[] {"#dt=${YEAR}", "${MONTH}", "${DAY}", "${HOUR}"}, separator);
String tableUri = "catalog:" + dbName + ":" + tblName + tableUriPartitionFragment;
final String datePattern = StringUtils.join(new String[] {"yyyy", "MM", "dd", "HH"}, separator);
// use the start date for both as this will only generate 2 partitions.
List<String> dataDates =
TimeUtil.getMinuteDatesOnEitherSide(
startDate, dataEndDate, 60, DateTimeFormat.forPattern(datePattern));
final List<String> dataset =
HadoopUtil.flattenAndPutDataInFolder(clusterFS, localHCatData, testHdfsDir, dataDates);
final String col1Name = "id";
final String col2Name = "value";
final String partitionColumn = "dt";
ArrayList<HCatFieldSchema> cols = new ArrayList<>();
cols.add(HCatUtil.getStringSchema(col1Name, col1Name + " comment"));
cols.add(HCatUtil.getStringSchema(col2Name, col2Name + " comment"));
ArrayList<HCatFieldSchema> partitionCols = new ArrayList<>();
// create table on cluster 1 and add data to it.
partitionCols.add(HCatUtil.getStringSchema(partitionColumn, partitionColumn + " partition"));
createTable(clusterHC, dbName, tblName, cols, partitionCols, testHdfsDir);
addPartitionsToTable(dataDates, dataset, "dt", dbName, tblName, clusterHC);
// create table on target cluster.
createTable(cluster2HC, dbName, tblName, cols, partitionCols, testHdfsDir);
createTable(cluster3HC, dbName, tblName, cols, partitionCols, testHdfsDir);
Bundle.submitCluster(bundles[0], bundles[1], bundles[2]);
bundles[0].setInputFeedPeriodicity(1, Frequency.TimeUnit.hours);
bundles[0].setInputFeedValidity(startDate, endDate);
bundles[0].setInputFeedTableUri(tableUri);
String feed = bundles[0].getDataSets().get(0);
// set the cluster 2 as the target.
feed =
FeedMerlin.fromString(feed)
.addFeedCluster(
new FeedMerlin.FeedClusterBuilder(
Util.readEntityName(bundles[1].getClusters().get(0)))
.withRetention("months(9000)", ActionType.DELETE)
.withValidity(startDate, endDate)
.withClusterType(ClusterType.TARGET)
.withTableUri(tableUri)
.build())
.toString();
// set the cluster 3 as the target.
feed =
FeedMerlin.fromString(feed)
.addFeedCluster(
new FeedMerlin.FeedClusterBuilder(
Util.readEntityName(bundles[2].getClusters().get(0)))
.withRetention("months(9000)", ActionType.DELETE)
.withValidity(startDate, endDate)
.withClusterType(ClusterType.TARGET)
.withTableUri(tableUri)
.build())
.toString();
AssertUtil.assertSucceeded(prism.getFeedHelper().submitAndSchedule(feed));
TimeUtil.sleepSeconds(TIMEOUT);
// check if all coordinators exist
Assert.assertEquals(
OozieUtil.checkIfFeedCoordExist(cluster2OC, Util.readEntityName(feed), "REPLICATION"), 1);
// check if all coordinators exist
Assert.assertEquals(
OozieUtil.checkIfFeedCoordExist(cluster3OC, Util.readEntityName(feed), "REPLICATION"), 1);
// replication should start, wait while it ends
// we will check for 2 instances so that both partitions are copied over.
InstanceUtil.waitTillInstanceReachState(
cluster2OC,
Util.readEntityName(feed),
2,
CoordinatorAction.Status.SUCCEEDED,
EntityType.FEED);
// replication should start, wait while it ends
// we will check for 2 instances so that both partitions are copied over.
InstanceUtil.waitTillInstanceReachState(
cluster3OC,
Util.readEntityName(feed),
2,
CoordinatorAction.Status.SUCCEEDED,
EntityType.FEED);
// check if data was replicated correctly
List<Path> srcData = HadoopUtil.getAllFilesRecursivelyHDFS(clusterFS, new Path(testHdfsDir));
LOGGER.info("Data on source cluster: " + srcData);
List<Path> cluster2TargetData =
HadoopUtil.getAllFilesRecursivelyHDFS(cluster2FS, new Path(testHdfsDir));
LOGGER.info("Data on target cluster: " + cluster2TargetData);
AssertUtil.checkForListSizes(srcData, cluster2TargetData);
List<Path> cluster3TargetData =
HadoopUtil.getAllFilesRecursivelyHDFS(cluster3FS, new Path(testHdfsDir));
LOGGER.info("Data on target cluster: " + cluster3TargetData);
AssertUtil.checkForListSizes(srcData, cluster3TargetData);
}
/**
* Set feed cluster1 as target, clusters 2 and 3 as source. Run feed. Update feed and check if
* action succeed. Check that appropriate number of replication and retention coordinators exist
* on matching clusters.
*
* @throws Exception
*/
@Test(enabled = true, timeOut = 1200000)
public void multipleSourceOneTarget() throws Exception {
bundles[0].setInputFeedDataPath(inputPath);
Bundle.submitCluster(bundles[0], bundles[1], bundles[2]);
FeedMerlin feed = new FeedMerlin(bundles[0].getDataSets().get(0));
feed.clearFeedClusters();
// use the colo string here so that the test works in embedded and distributed mode.
String postFix = "/US/" + cluster2Colo;
String prefix = bundles[0].getFeedDataPathPrefix();
HadoopUtil.deleteDirIfExists(prefix.substring(1), cluster2FS);
HadoopUtil.lateDataReplenish(cluster2FS, 5, 80, prefix, postFix);
// use the colo string here so that the test works in embedded and distributed mode.
postFix = "/UK/" + cluster3Colo;
prefix = bundles[0].getFeedDataPathPrefix();
HadoopUtil.deleteDirIfExists(prefix.substring(1), cluster3FS);
HadoopUtil.lateDataReplenish(cluster3FS, 5, 80, prefix, postFix);
String startTime = TimeUtil.getTimeWrtSystemTime(-30);
feed.addFeedCluster(
new FeedMerlin.FeedClusterBuilder(Util.readEntityName(bundles[1].getClusters().get(0)))
.withRetention("hours(10)", ActionType.DELETE)
.withValidity(startTime, TimeUtil.addMinsToTime(startTime, 85))
.withClusterType(ClusterType.SOURCE)
.withPartition("US/${cluster.colo}")
.build());
feed.addFeedCluster(
new FeedMerlin.FeedClusterBuilder(Util.readEntityName(bundles[0].getClusters().get(0)))
.withRetention("hours(10)", ActionType.DELETE)
.withValidity(
TimeUtil.addMinsToTime(startTime, 20), TimeUtil.addMinsToTime(startTime, 105))
.withClusterType(ClusterType.TARGET)
.build());
feed.addFeedCluster(
new FeedMerlin.FeedClusterBuilder(Util.readEntityName(bundles[2].getClusters().get(0)))
.withRetention("hours(10)", ActionType.DELETE)
.withValidity(
TimeUtil.addMinsToTime(startTime, 40), TimeUtil.addMinsToTime(startTime, 130))
.withClusterType(ClusterType.SOURCE)
.withPartition("UK/${cluster.colo}")
.build());
LOGGER.info("feed: " + Util.prettyPrintXml(feed.toString()));
AssertUtil.assertSucceeded(prism.getFeedHelper().submitEntity(feed.toString()));
AssertUtil.assertSucceeded(prism.getFeedHelper().schedule(feed.toString()));
// change feed location path
feed.setFilePath(alternativeInputPath);
LOGGER.info("updated feed: " + Util.prettyPrintXml(feed.toString()));
// update feed
AssertUtil.assertSucceeded(prism.getFeedHelper().update(feed.toString(), feed.toString()));
Assert.assertEquals(
OozieUtil.checkIfFeedCoordExist(cluster2OC, feed.getName(), "REPLICATION"), 0);
Assert.assertEquals(
OozieUtil.checkIfFeedCoordExist(cluster2OC, feed.getName(), "RETENTION"), 2);
Assert.assertEquals(
OozieUtil.checkIfFeedCoordExist(cluster3OC, feed.getName(), "REPLICATION"), 0);
Assert.assertEquals(
OozieUtil.checkIfFeedCoordExist(cluster3OC, feed.getName(), "RETENTION"), 2);
Assert.assertEquals(
OozieUtil.checkIfFeedCoordExist(cluster1OC, feed.getName(), "REPLICATION"), 4);
Assert.assertEquals(
OozieUtil.checkIfFeedCoordExist(cluster1OC, feed.getName(), "RETENTION"), 2);
}
