use of io.cdap.cdap.api.metadata.MetadataScope.USER in project cdap by caskdata.
the class MetadataStorageTest method testBatchConcurrency.
@Test
public void testBatchConcurrency() throws IOException {
// T threads
int numThreads = 10;
// N entities
int numEntities = 20;
MetadataStorage mds = getMetadataStorage();
ExecutorService executor = Executors.newFixedThreadPool(numThreads);
CompletionService<List<MetadataChange>> completionService = new ExecutorCompletionService<>(executor);
// Set up N entities with T tags (one to be removed per thread)
Set<String> allRTags = IntStream.range(0, numThreads).mapToObj(t -> "r" + t).collect(Collectors.toSet());
Map<Integer, MetadataEntity> entities = IntStream.range(0, numEntities).boxed().collect(Collectors.toMap(i -> i, i -> MetadataEntity.ofDataset("myds" + i)));
mds.batch(entities.values().stream().map(entity -> new Update(entity, new Metadata(USER, allRTags))).collect(Collectors.toList()), MutationOptions.DEFAULT);
// Generate a random but conflicting set of batches of mutations, one for each thread.
// Construct the expected results for each entity along with the mutations
// Also, because some threads will perform a Create, create a set of directives to preserve all other tags
Map<Integer, Set<String>> expected = IntStream.range(0, numEntities).boxed().collect(Collectors.toMap(i -> i, i -> new HashSet<>(allRTags)));
Map<Integer, List<MetadataMutation>> mutations = IntStream.range(0, numThreads).boxed().collect(Collectors.toMap(i -> i, i -> new ArrayList<>()));
Map<ScopedNameOfKind, MetadataDirective> directives = new HashMap<>();
Random rand = new Random(System.currentTimeMillis());
IntStream.range(0, numThreads).forEach(t -> {
directives.put(new ScopedNameOfKind(TAG, USER, "t" + t), MetadataDirective.KEEP);
directives.put(new ScopedNameOfKind(TAG, USER, "r" + t), MetadataDirective.KEEP);
directives.put(new ScopedNameOfKind(TAG, USER, "c" + t), MetadataDirective.KEEP);
IntStream.range(0, numEntities).forEach(e -> {
int random = rand.nextInt(100);
if (random < 30) {
expected.get(e).add("t" + t);
mutations.get(t).add(new Update(entities.get(e), new Metadata(USER, tags("t" + t))));
} else if (random < 60) {
expected.get(e).add("c" + t);
mutations.get(t).add(new Create(entities.get(e), new Metadata(USER, tags("c" + t)), directives));
} else if (random < 90) {
expected.get(e).remove("r" + t);
mutations.get(t).add(new Remove(entities.get(e), Collections.singleton(new ScopedNameOfKind(TAG, USER, "r" + t))));
}
});
});
// submit all tasks and wait for their completion
IntStream.range(0, numThreads).forEach(t -> completionService.submit(() -> mds.batch(mutations.get(t), MutationOptions.DEFAULT)));
IntStream.range(0, numThreads).forEach(t -> {
try {
completionService.take();
} catch (InterruptedException e) {
throw Throwables.propagate(e);
}
});
// validate that all "r" tags were removed and all "c" and "t" tags were added
IntStream.range(0, numEntities).forEach(e -> {
try {
Assert.assertEquals("For entity " + entities.get(e), expected.get(e), mds.read(new Read(entities.get(e))).getTags(USER));
} catch (Exception ex) {
throw Throwables.propagate(ex);
}
});
// clean up
mds.batch(entities.values().stream().map(Drop::new).collect(Collectors.toList()), MutationOptions.DEFAULT);
}
use of io.cdap.cdap.api.metadata.MetadataScope.USER in project cdap by caskdata.
the class MetadataStorageTest method testUpdateDropConflictInBatch.
/**
* See {@link #testUpdateDropConflict()} for a description. The difference in this test is that
* we issue batches of mutations over a collection of entities. The same assumptions apply,
* however, for each entity.
*/
@Test
@Category(SlowTests.class)
public void testUpdateDropConflictInBatch() throws IOException {
int numTests = 10;
int numThreads = 2;
int numEntities = 20;
MetadataStorage mds = getMetadataStorage();
ExecutorService executor = Executors.newFixedThreadPool(numThreads);
CompletionService<List<MetadataChange>> completionService = new ExecutorCompletionService<>(executor);
Map<Integer, MetadataEntity> entities = IntStream.range(0, numEntities).boxed().collect(Collectors.toMap(i -> i, i -> MetadataEntity.ofDataset("myds" + i)));
List<MetadataMutation> creates = entities.values().stream().map(e -> new Create(e, new Metadata(USER, tags("a")), Collections.emptyMap())).collect(Collectors.toList());
Random rand = new Random(System.currentTimeMillis());
IntStream.range(0, numTests).forEach(x -> {
try {
mds.batch(creates, MutationOptions.DEFAULT);
Map<Integer, List<MetadataMutation>> mutations = IntStream.range(0, numThreads).boxed().collect(Collectors.toMap(i -> i, i -> new ArrayList<>()));
IntStream.range(0, numEntities).forEach(e -> {
// ensure that at least one thread attempts to drop this entity
int dropThread = rand.nextInt(numThreads);
IntStream.range(0, numThreads).forEach(t -> {
if (t == dropThread || rand.nextInt(100) < 50) {
mutations.get(t).add(new Drop(entities.get(e)));
} else {
mutations.get(t).add(new Update(entities.get(e), new Metadata(USER, tags("b"))));
}
});
});
IntStream.range(0, numThreads).forEach(t -> completionService.submit(() -> mds.batch(mutations.get(t), MutationOptions.DEFAULT)));
IntStream.range(0, numThreads).forEach(t -> {
try {
completionService.take();
} catch (InterruptedException e) {
throw Throwables.propagate(e);
}
});
IntStream.range(0, numEntities).forEach(e -> {
try {
// each entity is either dropped then updated (and then it has tag "b" only)
// or it first update and then dropped (and then it has empty metadata)
Assert.assertTrue(ImmutableSet.of(Metadata.EMPTY, new Metadata(USER, tags("b"))).contains(mds.read(new Read(entities.get(e)))));
} catch (Exception ex) {
throw Throwables.propagate(ex);
}
});
} catch (Exception e) {
throw Throwables.propagate(e);
}
});
mds.batch(entities.values().stream().map(Drop::new).collect(Collectors.toList()), MutationOptions.DEFAULT);
}
use of io.cdap.cdap.api.metadata.MetadataScope.USER in project cdap by caskdata.
the class MetadataStorageTest method testBatch.
@Test
public void testBatch() throws IOException {
MetadataEntity entity = MetadataEntity.ofDataset("a", "b");
Map<ScopedNameOfKind, MetadataDirective> directives = ImmutableMap.of(new ScopedNameOfKind(PROPERTY, SYSTEM, CREATION_TIME_KEY), MetadataDirective.PRESERVE, new ScopedNameOfKind(PROPERTY, SYSTEM, DESCRIPTION_KEY), MetadataDirective.KEEP);
MetadataStorage mds = getMetadataStorage();
Create create = new Create(entity, new Metadata(SYSTEM, tags("batch"), props(CREATION_TIME_KEY, "12345678", DESCRIPTION_KEY, "hello", "other", "value")), directives);
MetadataChange change = mds.apply(create, MutationOptions.DEFAULT);
Assert.assertEquals(Metadata.EMPTY, change.getBefore());
Assert.assertEquals(create.getMetadata(), change.getAfter());
List<MetadataMutation> mutations = ImmutableList.of(new Update(entity, new Metadata(USER, tags("tag1", "tag2"))), new Drop(entity), new Create(entity, new Metadata(SYSTEM, tags("batch"), props(CREATION_TIME_KEY, "23456789", "other", "different")), directives), new Update(entity, new Metadata(USER, tags("tag3"), props("key", "value"))), new Remove(entity, ImmutableSet.of(new ScopedNameOfKind(PROPERTY, SYSTEM, "other"), new ScopedNameOfKind(TAG, USER, "tag2"))), new Create(entity, new Metadata(SYSTEM, tags("realtime"), props(CREATION_TIME_KEY, "33456789", DESCRIPTION_KEY, "new description", "other", "yet other")), directives));
// apply all mutations in sequence
List<MetadataChange> changes = mutations.stream().map(mutation -> {
try {
return mds.apply(mutation, MutationOptions.DEFAULT);
} catch (IOException e) {
throw Throwables.propagate(e);
}
}).collect(Collectors.toList());
// drop and recreate the entity
mds.apply(new Drop(entity), MutationOptions.DEFAULT);
change = mds.apply(create, MutationOptions.DEFAULT);
Assert.assertEquals(Metadata.EMPTY, change.getBefore());
Assert.assertEquals(create.getMetadata(), change.getAfter());
// apply all mutations in batch
List<MetadataChange> batchChanges = mds.batch(mutations, MutationOptions.DEFAULT);
// make sure the same mutations were applied
Assert.assertEquals(changes, batchChanges);
// clean up
mds.apply(new Drop(entity), MutationOptions.DEFAULT);
}
use of io.cdap.cdap.api.metadata.MetadataScope.USER in project cdap by caskdata.
the class MetadataStorageTest method testSearchPagination.
@Test
public void testSearchPagination() throws IOException {
MetadataStorage mds = getMetadataStorage();
String ns = "ns";
MetadataEntity app1 = ofApp(ns, "app");
MetadataEntity app2 = ofApp(ns, "app2");
MetadataEntity service = ofService(app1, "service");
MetadataEntity worker = ofWorker(app2, "worker");
MetadataEntity dataset = ofDataset(ns, "dataset");
MetadataEntity hidden = ofDataset(ns, "_auditLog");
MetadataRecord serviceRecord = new MetadataRecord(service, new Metadata(USER, tags("tag", "tag1")));
MetadataRecord workerRecord = new MetadataRecord(worker, new Metadata(USER, tags("tag2", "tag3 tag4")));
MetadataRecord datasetRecord = new MetadataRecord(dataset, new Metadata(USER, tags("tag5 tag6", "tag7 tag8")));
MetadataRecord hiddenRecord = new MetadataRecord(hidden, new Metadata(USER, tags("tag9", "tag10", "tag11", "tag12", "tag13")));
mds.batch(batch(new Update(service, serviceRecord.getMetadata()), new Update(worker, workerRecord.getMetadata()), new Update(dataset, datasetRecord.getMetadata()), new Update(hidden, hiddenRecord.getMetadata())), MutationOptions.DEFAULT);
// assert that search returns all records and determine the order of relevance )it varies by implementation)
SearchResponse response = mds.search(SearchRequest.of("tag*").addNamespace(ns).setShowHidden(true).setLimit(Integer.MAX_VALUE).build());
List<MetadataRecord> results = response.getResults();
List<MetadataRecord> noHidden = results.stream().filter(record -> !record.equals(hiddenRecord)).collect(Collectors.toList());
Assert.assertEquals(ImmutableSet.of(serviceRecord, workerRecord, datasetRecord, hiddenRecord), ImmutableSet.copyOf(results));
assertResults(mds, SearchRequest.of("tag*").addNamespace(ns).setLimit(Integer.MAX_VALUE).build(), noHidden.get(0), noHidden.get(1), noHidden.get(2));
// hidden entity should now be returned since showHidden is true
assertResults(mds, SearchRequest.of("tag*").addNamespace(ns).setShowHidden(true).build(), results.get(0), results.get(1), results.get(2), results.get(3));
assertResults(mds, SearchRequest.of("tag*").addNamespace(ns).setLimit(2).setCursorRequested(true).build(), noHidden.get(0), noHidden.get(1));
// skipping hidden entity should not affect the offset
assertResults(mds, SearchRequest.of("tag*").addNamespace(ns).setOffset(1).setLimit(2).build(), noHidden.get(1), noHidden.get(2));
// if showHidden is true, the hidden entity should affect the offset
assertResults(mds, SearchRequest.of("tag*").addNamespace(ns).setOffset(1).setLimit(3).setShowHidden(true).build(), results.get(1), results.get(2), results.get(3));
assertResults(mds, SearchRequest.of("tag*").addNamespace(ns).setOffset(2).setLimit(2).build(), noHidden.get(2));
assertEmpty(mds, SearchRequest.of("tag*").addNamespace(ns).setOffset(4).setLimit(2).build());
assertResults(mds, SearchRequest.of("tag*").addNamespace(ns).setOffset(1).build(), noHidden.get(1), noHidden.get(2));
// clean up
mds.batch(batch(new Drop(service), new Drop(worker), new Drop(dataset), new Drop(hidden)), MutationOptions.DEFAULT);
}
use of io.cdap.cdap.api.metadata.MetadataScope.USER in project cdap by caskdata.
the class DatasetMetadataStorageTest method testSearchWeight.
// this tests is not in MetadataStorageTest,
// because it tests result scoring and sorting specific to the dataset-based implementation
@Test
public void testSearchWeight() throws IOException {
MetadataStorage mds = getMetadataStorage();
String ns = "ns1";
NamespaceId nsId = new NamespaceId(ns);
MetadataEntity service1 = nsId.app("app1").service("service1").toMetadataEntity();
MetadataEntity dataset1 = nsId.dataset("ds1").toMetadataEntity();
MetadataEntity dataset2 = nsId.dataset("ds2").toMetadataEntity();
// Add metadata
String multiWordValue = "aV1 av2 , - , av3 - av4_av5 av6";
Map<String, String> userProps = ImmutableMap.of("key1", "value1", "key2", "value2", "multiword", multiWordValue);
Map<String, String> systemProps = ImmutableMap.of("sysKey1", "sysValue1");
Set<String> userTags = ImmutableSet.of("tag1", "tag2");
Set<String> temporaryUserTags = ImmutableSet.of("tag3", "tag4");
Map<String, String> dataset1UserProps = ImmutableMap.of("sKey1", "sValuee1 sValuee2");
Map<String, String> dataset2UserProps = ImmutableMap.of("sKey1", "sValue1 sValue2", "Key1", "Value1");
Set<String> sysTags = ImmutableSet.of("sysTag1");
MetadataRecord service1Record = new MetadataRecord(service1, union(new Metadata(USER, userTags, userProps), new Metadata(SYSTEM, sysTags, systemProps)));
mds.apply(new Update(service1Record.getEntity(), service1Record.getMetadata()), MutationOptions.DEFAULT);
// dd and then remove some metadata for dataset2
mds.apply(new Update(dataset2, new Metadata(USER, temporaryUserTags, userProps)), MutationOptions.DEFAULT);
mds.apply(new Remove(dataset2, temporaryUserTags.stream().map(tag -> new ScopedNameOfKind(TAG, USER, tag)).collect(Collectors.toSet())), MutationOptions.DEFAULT);
mds.apply(new Remove(dataset2, userProps.keySet().stream().map(tag -> new ScopedNameOfKind(PROPERTY, USER, tag)).collect(Collectors.toSet())), MutationOptions.DEFAULT);
MetadataRecord dataset1Record = new MetadataRecord(dataset1, new Metadata(USER, tags(), dataset1UserProps));
MetadataRecord dataset2Record = new MetadataRecord(dataset2, new Metadata(USER, tags(), dataset2UserProps));
mds.batch(ImmutableList.of(new Update(dataset1Record.getEntity(), dataset1Record.getMetadata()), new Update(dataset2Record.getEntity(), dataset2Record.getMetadata())), MutationOptions.DEFAULT);
// Test score and metadata match
assertInOrder(mds, SearchRequest.of("value1 multiword:av2").addNamespace(ns).build(), service1Record, dataset2Record);
assertInOrder(mds, SearchRequest.of("value1 sValue*").addNamespace(ns).setLimit(Integer.MAX_VALUE).build(), dataset2Record, dataset1Record, service1Record);
assertResults(mds, SearchRequest.of("*").addNamespace(ns).setLimit(Integer.MAX_VALUE).build(), dataset2Record, dataset1Record, service1Record);
// clean up
mds.batch(ImmutableList.of(new Drop(service1), new Drop(dataset1), new Drop(dataset2)), MutationOptions.DEFAULT);
}
Aggregations