use of org.apache.hudi.client.SparkRDDWriteClient in project hudi by apache.
the class TestInlineCompaction method testSuccessfulCompactionBasedOnNumCommits.
@Test
public void testSuccessfulCompactionBasedOnNumCommits() throws Exception {
// Given: make three commits
HoodieWriteConfig cfg = getConfigForInlineCompaction(3, 60, CompactionTriggerStrategy.NUM_COMMITS);
List<String> instants = IntStream.range(0, 2).mapToObj(i -> HoodieActiveTimeline.createNewInstantTime()).collect(Collectors.toList());
try (SparkRDDWriteClient<?> writeClient = getHoodieWriteClient(cfg)) {
List<HoodieRecord> records = dataGen.generateInserts(instants.get(0), 100);
HoodieReadClient readClient = getHoodieReadClient(cfg.getBasePath());
runNextDeltaCommits(writeClient, readClient, instants, records, cfg, true, new ArrayList<>());
// third commit, that will trigger compaction
HoodieTableMetaClient metaClient = HoodieTableMetaClient.builder().setConf(hadoopConf).setBasePath(cfg.getBasePath()).build();
String finalInstant = HoodieActiveTimeline.createNewInstantTime();
createNextDeltaCommit(finalInstant, dataGen.generateUpdates(finalInstant, 100), writeClient, metaClient, cfg, false);
// Then: ensure the file slices are compacted as per policy
metaClient = HoodieTableMetaClient.builder().setConf(hadoopConf).setBasePath(cfg.getBasePath()).build();
assertEquals(4, metaClient.getActiveTimeline().getWriteTimeline().countInstants());
assertEquals(HoodieTimeline.COMMIT_ACTION, metaClient.getActiveTimeline().lastInstant().get().getAction());
String compactionTime = metaClient.getActiveTimeline().lastInstant().get().getTimestamp();
assertFalse(WriteMarkersFactory.get(cfg.getMarkersType(), HoodieSparkTable.create(cfg, context), compactionTime).doesMarkerDirExist());
}
}
use of org.apache.hudi.client.SparkRDDWriteClient in project hudi by apache.
the class TestInlineCompaction method testSuccessfulCompactionBasedOnNumOrTime.
@Test
public void testSuccessfulCompactionBasedOnNumOrTime() throws Exception {
// Given: make three commits
HoodieWriteConfig cfg = getConfigForInlineCompaction(3, 20, CompactionTriggerStrategy.NUM_OR_TIME);
try (SparkRDDWriteClient<?> writeClient = getHoodieWriteClient(cfg)) {
List<HoodieRecord> records = dataGen.generateInserts(HoodieActiveTimeline.createNewInstantTime(), 10);
HoodieReadClient readClient = getHoodieReadClient(cfg.getBasePath());
List<String> instants = IntStream.range(0, 2).mapToObj(i -> HoodieActiveTimeline.createNewInstantTime()).collect(Collectors.toList());
runNextDeltaCommits(writeClient, readClient, instants, records, cfg, true, new ArrayList<>());
// Then: trigger the compaction because reach 3 commits.
String finalInstant = HoodieActiveTimeline.createNewInstantTime();
HoodieTableMetaClient metaClient = HoodieTableMetaClient.builder().setConf(hadoopConf).setBasePath(cfg.getBasePath()).build();
createNextDeltaCommit(finalInstant, dataGen.generateUpdates(finalInstant, 10), writeClient, metaClient, cfg, false);
metaClient = HoodieTableMetaClient.builder().setConf(hadoopConf).setBasePath(cfg.getBasePath()).build();
assertEquals(4, metaClient.getActiveTimeline().getWriteTimeline().countInstants());
// 4th commit, that will trigger compaction because reach the time elapsed
metaClient = HoodieTableMetaClient.builder().setConf(hadoopConf).setBasePath(cfg.getBasePath()).build();
finalInstant = HoodieActiveTimeline.createNewInstantTime(20000);
createNextDeltaCommit(finalInstant, dataGen.generateUpdates(finalInstant, 10), writeClient, metaClient, cfg, false);
metaClient = HoodieTableMetaClient.builder().setConf(hadoopConf).setBasePath(cfg.getBasePath()).build();
assertEquals(6, metaClient.getActiveTimeline().getWriteTimeline().countInstants());
}
}
use of org.apache.hudi.client.SparkRDDWriteClient in project hudi by apache.
the class TestInlineCompaction method testCompactionRetryOnFailureBasedOnTime.
@Test
public void testCompactionRetryOnFailureBasedOnTime() throws Exception {
// Given: two commits, schedule compaction and its failed/in-flight
HoodieWriteConfig cfg = getConfigBuilder(false).withCompactionConfig(HoodieCompactionConfig.newBuilder().withInlineCompaction(false).withMaxDeltaSecondsBeforeCompaction(5).withInlineCompactionTriggerStrategy(CompactionTriggerStrategy.TIME_ELAPSED).build()).build();
String instantTime;
List<String> instants = IntStream.range(0, 2).mapToObj(i -> HoodieActiveTimeline.createNewInstantTime()).collect(Collectors.toList());
try (SparkRDDWriteClient<?> writeClient = getHoodieWriteClient(cfg)) {
List<HoodieRecord> records = dataGen.generateInserts(instants.get(0), 100);
HoodieReadClient readClient = getHoodieReadClient(cfg.getBasePath());
runNextDeltaCommits(writeClient, readClient, instants, records, cfg, true, new ArrayList<>());
// Schedule compaction instantTime, make it in-flight (simulates inline compaction failing)
instantTime = HoodieActiveTimeline.createNewInstantTime(10000);
scheduleCompaction(instantTime, writeClient, cfg);
moveCompactionFromRequestedToInflight(instantTime, cfg);
}
// When: commit happens after 10s
HoodieWriteConfig inlineCfg = getConfigForInlineCompaction(5, 10, CompactionTriggerStrategy.TIME_ELAPSED);
String instantTime2;
try (SparkRDDWriteClient<?> writeClient = getHoodieWriteClient(inlineCfg)) {
HoodieTableMetaClient metaClient = HoodieTableMetaClient.builder().setConf(hadoopConf).setBasePath(cfg.getBasePath()).build();
instantTime2 = HoodieActiveTimeline.createNewInstantTime();
createNextDeltaCommit(instantTime2, dataGen.generateUpdates(instantTime2, 10), writeClient, metaClient, inlineCfg, false);
}
// Then: 1 delta commit is done, the failed compaction is retried
metaClient = HoodieTableMetaClient.builder().setConf(hadoopConf).setBasePath(cfg.getBasePath()).build();
assertEquals(4, metaClient.getActiveTimeline().getWriteTimeline().countInstants());
assertEquals(instantTime, metaClient.getActiveTimeline().getCommitTimeline().filterCompletedInstants().firstInstant().get().getTimestamp());
}
use of org.apache.hudi.client.SparkRDDWriteClient in project hudi by apache.
the class TestInlineCompaction method testCompactionIsNotScheduledEarly.
@Test
public void testCompactionIsNotScheduledEarly() throws Exception {
// Given: make two commits
HoodieWriteConfig cfg = getConfigForInlineCompaction(3, 60, CompactionTriggerStrategy.NUM_COMMITS);
try (SparkRDDWriteClient<?> writeClient = getHoodieWriteClient(cfg)) {
List<HoodieRecord> records = dataGen.generateInserts(HoodieActiveTimeline.createNewInstantTime(), 100);
HoodieReadClient readClient = getHoodieReadClient(cfg.getBasePath());
List<String> instants = IntStream.range(0, 2).mapToObj(i -> HoodieActiveTimeline.createNewInstantTime()).collect(Collectors.toList());
runNextDeltaCommits(writeClient, readClient, instants, records, cfg, true, new ArrayList<>());
HoodieTableMetaClient metaClient = HoodieTableMetaClient.builder().setConf(hadoopConf).setBasePath(cfg.getBasePath()).build();
// Then: ensure no compaction is executed since there are only 2 delta commits
assertEquals(2, metaClient.getActiveTimeline().getWriteTimeline().countInstants());
}
}
use of org.apache.hudi.client.SparkRDDWriteClient in project hudi by apache.
the class TestMergeOnReadRollbackActionExecutor method testRollbackWhenFirstCommitFail.
/**
* Test Cases for rolling back when there is no base file.
*/
@Test
public void testRollbackWhenFirstCommitFail() throws Exception {
HoodieWriteConfig config = HoodieWriteConfig.newBuilder().withRollbackUsingMarkers(false).withPath(basePath).build();
try (SparkRDDWriteClient client = getHoodieWriteClient(config)) {
client.startCommitWithTime("001");
client.insert(jsc.emptyRDD(), "001");
client.rollback("001");
}
}
Aggregations