honeymoose
/
OpenSearch
mirror of https://github.com/honeymoose/OpenSearch.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Batch processing mapping updates can cause missed merged mappings when batching multiple types 

when we bulk changes, we need to use the same index metadata builder across the tasks, otherwise we might remove mappings erroneously
 also, when we check if we can use a higher order mapping, we need to verify that its for the same mapping type
This commit is contained in:
Shay Banon 2013-12-11 14:34:50 +01:00
parent a760f1f54a
commit 10cdb0ae22
4 changed files with 162 additions and 201 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
src/main/java/org/elasticsearch/cluster/metadata/IndexMetaData.java
View File

@ -466,6 +466,10 @@ public class IndexMetaData {
return this;
}
public MappingMetaData mapping(String type) {
return mappings.get(type);
}
public Builder removeMapping(String mappingType) {
mappings.remove(mappingType);
return this;

209
src/main/java/org/elasticsearch/cluster/metadata/MetaDataMappingService.java
View File

@ -161,9 +161,10 @@ public class MetaDataMappingService extends AbstractComponent {
boolean dirty = false;
MetaData.Builder mdBuilder = MetaData.builder(currentState.metaData());
for (Map.Entry<String, List<MappingTask>> entry : tasksPerIndex.entrySet()) {
String index = entry.getKey();
final IndexMetaData indexMetaData = mdBuilder.get(index);
IndexMetaData indexMetaData = mdBuilder.get(index);
if (indexMetaData == null) {
// index got deleted on us, ignore...
logger.debug("[{}] ignoring tasks - index meta data doesn't exist", index);
@ -188,13 +189,15 @@ public class MetaDataMappingService extends AbstractComponent {
MappingTask existing = tasks.get(i);
if (existing instanceof UpdateTask) {
UpdateTask eTask = (UpdateTask) existing;
// if we have the order, and the node id, then we can compare, and replace if applicable
if (eTask.order != -1 && eTask.nodeId != null) {
if (eTask.nodeId.equals(uTask.nodeId) && uTask.order > eTask.order) {
// a newer update task, we can replace so we execute it one!
tasks.set(i, uTask);
add = false;
break;
if (eTask.type.equals(uTask.type)) {
// if we have the order, and the node id, then we can compare, and replace if applicable
if (eTask.order != -1 && eTask.nodeId != null) {
if (eTask.nodeId.equals(uTask.nodeId) && uTask.order > eTask.order) {
// a newer update task, we can replace so we execute it one!
tasks.set(i, uTask);
add = false;
break;
}
}
}
}
@ -207,98 +210,37 @@ public class MetaDataMappingService extends AbstractComponent {
}
}
// construct the actual index if needed, and make sure the relevant mappings are there
boolean removeIndex = false;
// keep track of what we already refreshed, no need to refresh it again...
Set<String> processedRefreshes = Sets.newHashSet();
try {
IndexService indexService = indicesService.indexService(index);
if (indexService == null) {
// we need to create the index here, and add the current mapping to it, so we can merge
indexService = indicesService.createIndex(indexMetaData.index(), indexMetaData.settings(), currentState.nodes().localNode().id());
removeIndex = true;
Set<String> typesToIntroduce = Sets.newHashSet();
for (MappingTask task : tasks) {
if (task instanceof RefreshTask) {
RefreshTask refreshTask = (RefreshTask) task;
try {
IndexService indexService = indicesService.indexService(index);
if (indexService == null) {
// we need to create the index here, and add the current mapping to it, so we can merge
indexService = indicesService.createIndex(indexMetaData.index(), indexMetaData.settings(), currentState.nodes().localNode().id());
removeIndex = true;
for (String type : refreshTask.types) {
// only add the current relevant mapping (if exists)
if (indexMetaData.mappings().containsKey(type)) {
// don't apply the default mapping, it has been applied when the mapping was created
indexService.mapperService().merge(type, indexMetaData.mappings().get(type).source(), false);
}
}
}
IndexMetaData.Builder indexMetaDataBuilder = IndexMetaData.builder(indexMetaData);
List<String> updatedTypes = Lists.newArrayList();
for (String type : refreshTask.types) {
if (processedRefreshes.contains(type)) {
continue;
}
DocumentMapper mapper = indexService.mapperService().documentMapper(type);
if (!mapper.mappingSource().equals(indexMetaData.mappings().get(type).source())) {
updatedTypes.add(type);
indexMetaDataBuilder.putMapping(new MappingMetaData(mapper));
}
processedRefreshes.add(type);
}
if (updatedTypes.isEmpty()) {
continue;
}
logger.warn("[{}] re-syncing mappings with cluster state for types [{}]", index, updatedTypes);
mdBuilder.put(indexMetaDataBuilder);
dirty = true;
} catch (Throwable t) {
logger.warn("[{}] failed to refresh-mapping in cluster state, types [{}]", index, refreshTask.types);
}
} else if (task instanceof UpdateTask) {
UpdateTask updateTask = (UpdateTask) task;
try {
String type = updateTask.type;
CompressedString mappingSource = updateTask.mappingSource;
if (indexMetaData.mappings().containsKey(type) && indexMetaData.mapping(type).source().equals(mappingSource)) {
logger.debug("[{}] update_mapping [{}] ignoring mapping update task as its source is equal to ours", index, updateTask.type);
continue;
}
IndexService indexService = indicesService.indexService(index);
if (indexService == null) {
// we need to create the index here, and add the current mapping to it, so we can merge
indexService = indicesService.createIndex(indexMetaData.index(), indexMetaData.settings(), currentState.nodes().localNode().id());
removeIndex = true;
// only add the current relevant mapping (if exists)
if (indexMetaData.mappings().containsKey(type)) {
indexService.mapperService().merge(type, indexMetaData.mappings().get(type).source(), false);
}
}
DocumentMapper updatedMapper = indexService.mapperService().merge(type, mappingSource, false);
processedRefreshes.add(type);
// if we end up with the same mapping as the original once, ignore
if (indexMetaData.mappings().containsKey(type) && indexMetaData.mapping(type).source().equals(updatedMapper.mappingSource())) {
logger.debug("[{}] update_mapping [{}] ignoring mapping update task as it results in the same source as what we have", index, updateTask.type);
continue;
}
// build the updated mapping source
if (logger.isDebugEnabled()) {
logger.debug("[{}] update_mapping [{}] (dynamic) with source [{}]", index, type, updatedMapper.mappingSource());
} else if (logger.isInfoEnabled()) {
logger.info("[{}] update_mapping [{}] (dynamic)", index, type);
}
mdBuilder.put(IndexMetaData.builder(indexMetaData).putMapping(new MappingMetaData(updatedMapper)));
dirty = true;
} catch (Throwable t) {
logger.warn("[{}] failed to update-mapping in cluster state, type [{}]", index, updateTask.type);
}
} else {
logger.warn("illegal state, got wrong mapping task type [{}]", task);
if (task instanceof UpdateTask) {
typesToIntroduce.add(((UpdateTask) task).type);
} else if (task instanceof RefreshTask) {
Collections.addAll(typesToIntroduce, ((RefreshTask) task).types);
}
}
for (String type : typesToIntroduce) {
// only add the current relevant mapping (if exists)
if (indexMetaData.mappings().containsKey(type)) {
// don't apply the default mapping, it has been applied when the mapping was created
indexService.mapperService().merge(type, indexMetaData.mappings().get(type).source(), false);
}
}
}
IndexMetaData.Builder builder = IndexMetaData.builder(indexMetaData);
try {
boolean indexDirty = processIndexMappingTasks(tasks, indexService, builder);
if (indexDirty) {
mdBuilder.put(builder);
dirty = true;
}
} finally {
if (removeIndex) {
indicesService.removeIndex(index, "created for mapping processing");
@ -321,13 +263,86 @@ public class MetaDataMappingService extends AbstractComponent {
}
});
if (!dirty) {
return currentState;
}
return ClusterState.builder(currentState).metaData(mdBuilder).build();
}
private boolean processIndexMappingTasks(List<MappingTask> tasks, IndexService indexService, IndexMetaData.Builder builder) {
boolean dirty = false;
String index = indexService.index().name();
// keep track of what we already refreshed, no need to refresh it again...
Set<String> processedRefreshes = Sets.newHashSet();
for (MappingTask task : tasks) {
if (task instanceof RefreshTask) {
RefreshTask refreshTask = (RefreshTask) task;
try {
List<String> updatedTypes = Lists.newArrayList();
for (String type : refreshTask.types) {
if (processedRefreshes.contains(type)) {
continue;
}
DocumentMapper mapper = indexService.mapperService().documentMapper(type);
if (mapper == null) {
continue;
}
if (!mapper.mappingSource().equals(builder.mapping(type).source())) {
updatedTypes.add(type);
builder.putMapping(new MappingMetaData(mapper));
}
processedRefreshes.add(type);
}
if (updatedTypes.isEmpty()) {
continue;
}
logger.warn("[{}] re-syncing mappings with cluster state for types [{}]", index, updatedTypes);
dirty = true;
} catch (Throwable t) {
logger.warn("[{}] failed to refresh-mapping in cluster state, types [{}]", index, refreshTask.types);
}
} else if (task instanceof UpdateTask) {
UpdateTask updateTask = (UpdateTask) task;
try {
String type = updateTask.type;
CompressedString mappingSource = updateTask.mappingSource;
MappingMetaData mappingMetaData = builder.mapping(type);
if (mappingMetaData != null && mappingMetaData.source().equals(mappingSource)) {
logger.debug("[{}] update_mapping [{}] ignoring mapping update task as its source is equal to ours", index, updateTask.type);
continue;
}
DocumentMapper updatedMapper = indexService.mapperService().merge(type, mappingSource, false);
processedRefreshes.add(type);
// if we end up with the same mapping as the original once, ignore
if (mappingMetaData != null && mappingMetaData.source().equals(updatedMapper.mappingSource())) {
logger.debug("[{}] update_mapping [{}] ignoring mapping update task as it results in the same source as what we have", index, updateTask.type);
continue;
}
// build the updated mapping source
if (logger.isDebugEnabled()) {
logger.debug("[{}] update_mapping [{}] (dynamic) with source [{}]", index, type, updatedMapper.mappingSource());
} else if (logger.isInfoEnabled()) {
logger.info("[{}] update_mapping [{}] (dynamic)", index, type);
}
builder.putMapping(new MappingMetaData(updatedMapper));
dirty = true;
} catch (Throwable t) {
logger.warn("[{}] failed to update-mapping in cluster state, type [{}]", index, updateTask.type);
}
} else {
logger.warn("illegal state, got wrong mapping task type [{}]", task);
}
}
return dirty;
}
/**
* Refreshes mappings if they are not the same between original and parsed version
*/

4
src/main/java/org/elasticsearch/index/mapper/DocumentMapper.java
View File

@ -655,13 +655,13 @@ public class DocumentMapper implements ToXContent {
return new MergeResult(mergeContext.buildConflicts());
}
public void refreshSource() throws FailedToGenerateSourceMapperException {
public CompressedString refreshSource() throws FailedToGenerateSourceMapperException {
try {
XContentBuilder builder = XContentFactory.contentBuilder(XContentType.JSON);
builder.startObject();
toXContent(builder, ToXContent.EMPTY_PARAMS);
builder.endObject();
this.mappingSource = new CompressedString(builder.bytes());
return mappingSource = new CompressedString(builder.bytes());
} catch (Exception e) {
throw new FailedToGenerateSourceMapperException(e.getMessage(), e);
}

146
src/test/java/org/elasticsearch/indices/mapping/UpdateMappingTests.java
View File

@ -1,19 +1,18 @@
package org.elasticsearch.indices.mapping;
import com.carrotsearch.hppc.cursors.ObjectObjectCursor;
import com.google.common.base.Predicate;
import org.elasticsearch.action.ActionListener;
import com.google.common.collect.Lists;
import com.google.common.collect.Maps;
import org.elasticsearch.action.admin.cluster.tasks.PendingClusterTasksResponse;
import org.elasticsearch.action.admin.indices.mapping.get.GetMappingsResponse;
import org.elasticsearch.action.admin.indices.mapping.put.PutMappingResponse;
import org.elasticsearch.action.admin.indices.refresh.RefreshResponse;
import org.elasticsearch.action.bulk.BulkRequestBuilder;
import org.elasticsearch.action.bulk.BulkResponse;
import org.elasticsearch.action.count.CountResponse;
import org.elasticsearch.action.get.GetResponse;
import org.elasticsearch.action.index.IndexRequest;
import org.elasticsearch.action.index.IndexRequestBuilder;
import org.elasticsearch.action.index.IndexResponse;
import org.elasticsearch.client.Client;
import org.elasticsearch.cluster.ClusterState;
import org.elasticsearch.cluster.metadata.MappingMetaData;
import org.elasticsearch.common.Priority;
import org.elasticsearch.common.collect.ImmutableOpenMap;
@ -23,13 +22,12 @@ import org.elasticsearch.index.mapper.MapperParsingException;
import org.elasticsearch.index.mapper.MapperService;
import org.elasticsearch.index.mapper.MergeMappingException;
import org.elasticsearch.test.ElasticsearchIntegrationTest;
import org.elasticsearch.threadpool.ThreadPool;
import org.hamcrest.Matchers;
import org.junit.Test;
import java.util.ArrayList;
import java.util.List;
import java.util.Map;
import java.util.concurrent.CountDownLatch;
import java.util.concurrent.CyclicBarrier;
import java.util.concurrent.atomic.AtomicBoolean;
@ -41,7 +39,7 @@ import static org.hamcrest.Matchers.*;
public class UpdateMappingTests extends ElasticsearchIntegrationTest {
@Test
public void dynamicUpdates_Index() throws Exception {
public void dynamicUpdates() throws Exception {
client().admin().indices().prepareCreate("test")
.setSettings(
ImmutableSettings.settingsBuilder()
@ -50,32 +48,15 @@ public class UpdateMappingTests extends ElasticsearchIntegrationTest {
).execute().actionGet();
client().admin().cluster().prepareHealth().setWaitForEvents(Priority.LANGUID).setWaitForGreenStatus().execute().actionGet();
ThreadPool.Info info = cluster().getInstance(ThreadPool.class).info(ThreadPool.Names.INDEX);
int recCount = info.getMax() + (int) info.getQueueSize().getSingles();
final CountDownLatch latch = new CountDownLatch(recCount);
int recCount = randomIntBetween(200, 600);
int numberOfTypes = randomIntBetween(1, 5);
List<IndexRequestBuilder> indexRequests = Lists.newArrayList();
for (int rec = 0; rec < recCount; rec++) {
client().prepareIndex("test", "type", "rec" + rec).setSource("field" + rec, "some_value").execute(new ActionListener<IndexResponse>() {
@Override
public void onResponse(IndexResponse indexResponse) {
latch.countDown();
}
@Override
public void onFailure(Throwable e) {
logger.error("failed to index in test", e);
latch.countDown();
}
});
String type = "type" + (rec % numberOfTypes);
String fieldName = "field_" + type + "_" + rec;
indexRequests.add(client().prepareIndex("test", type, Integer.toString(rec)).setSource(fieldName, "some_value"));
}
latch.await();
logger.info("wait till the mappings have been processed...");
awaitBusy(new Predicate<Object>() {
@Override
public boolean apply(Object input) {
PendingClusterTasksResponse pendingTasks = client().admin().cluster().preparePendingClusterTasks().get();
return pendingTasks.pendingTasks().isEmpty();
}
});
indexRandom(true, indexRequests);
logger.info("checking all the documents are there");
RefreshResponse refreshResponse = client().admin().indices().prepareRefresh().execute().actionGet();
@ -84,76 +65,37 @@ public class UpdateMappingTests extends ElasticsearchIntegrationTest {
assertThat(response.getCount(), equalTo((long) recCount));
logger.info("checking all the fields are in the mappings");
String source = client().admin().cluster().prepareState().get().getState().getMetaData().getIndices().get("test").getMappings().get("type").source().string();
for (int rec = 0; rec < recCount; rec++) {
assertThat(source, containsString("\"field" + rec + "\""));
}
}
@Test
public void dynamicUpdates_Bulk() throws Exception {
client().admin().indices().prepareCreate("test")
.setSettings(
ImmutableSettings.settingsBuilder()
.put("index.number_of_shards", 1)
.put("index.number_of_replicas", 0)
).execute().actionGet();
client().admin().cluster().prepareHealth().setWaitForEvents(Priority.LANGUID).setWaitForGreenStatus().execute().actionGet();
ThreadPool.Info info = cluster().getInstance(ThreadPool.class).info(ThreadPool.Names.BULK);
int bulkCount = info.getMax() + (int) info.getQueueSize().getSingles();
int bulkSize = between(4, 10);
int recCount = bulkCount * bulkSize;
int idCounter = 0;
final CountDownLatch latch = new CountDownLatch(bulkCount);
for (int i = 0; i < bulkCount; i++) {
BulkRequestBuilder bulk = client().prepareBulk();
for (int rec = 0; rec < bulkSize; rec++) {
int id = idCounter++;
bulk.add(new IndexRequestBuilder(client())
.setOpType(IndexRequest.OpType.INDEX)
.setIndex("test")
.setType("type")
.setId("rec" + id)
.setSource("field" + id, "some_value"));
reRunTest:
while (true) {
Map<String, String> typeToSource = Maps.newHashMap();
ClusterState state = client().admin().cluster().prepareState().get().getState();
for (ObjectObjectCursor<String, MappingMetaData> cursor : state.getMetaData().getIndices().get("test").getMappings()) {
typeToSource.put(cursor.key, cursor.value.source().string());
}
bulk.execute(new ActionListener<BulkResponse>() {
@Override
public void onResponse(BulkResponse bulkItemResponses) {
if (bulkItemResponses.hasFailures()) {
System.out.println("failed to index in test: " + bulkItemResponses.buildFailureMessage());
for (int rec = 0; rec < recCount; rec++) {
String type = "type" + (rec % numberOfTypes);
String fieldName = "field_" + type + "_" + rec;
fieldName = "\"" + fieldName + "\""; // quote it, so we make sure we catch the exact one
if (!typeToSource.containsKey(type) || !typeToSource.get(type).contains(fieldName)) {
awaitBusy(new Predicate<Object>() {
@Override
public boolean apply(Object input) {
PendingClusterTasksResponse pendingTasks = client().admin().cluster().preparePendingClusterTasks().get();
return pendingTasks.pendingTasks().isEmpty();
}
});
// its going to break, before we do, make sure that the cluster state hasn't changed on us...
ClusterState state2 = client().admin().cluster().prepareState().get().getState();
if (state.version() != state2.version()) {
logger.info("not the same version, used for test {}, new one {}, re-running test, first wait for mapping to wait", state.version(), state2.version());
continue reRunTest;
}
latch.countDown();
logger.info("failing, type {}, field {}, mapping {}", type, fieldName, typeToSource.get(type));
assertThat(typeToSource.get(type), containsString(fieldName));
}
@Override
public void onFailure(Throwable e) {
logger.error("failed to index in test", e);
latch.countDown();
}
});
}
latch.await();
logger.info("wait till the mappings have been processed...");
awaitBusy(new Predicate<Object>() {
@Override
public boolean apply(Object input) {
PendingClusterTasksResponse pendingTasks = client().admin().cluster().preparePendingClusterTasks().get();
return pendingTasks.pendingTasks().isEmpty();
}
});
logger.info("checking all the documents are there");
RefreshResponse refreshResponse = client().admin().indices().prepareRefresh().execute().actionGet();
assertThat(refreshResponse.getFailedShards(), equalTo(0));
CountResponse response = client().prepareCount("test").execute().actionGet();
assertThat(response.getCount(), equalTo((long) recCount));
logger.info("checking all the fields are in the mappings");
String source = client().admin().cluster().prepareState().get().getState().getMetaData().getIndices().get("test").getMappings().get("type").source().string();
for (int rec = 0; rec < recCount; rec++) {
assertThat(source, containsString("\"field" + rec + "\""));
break;
}
}
@ -227,12 +169,12 @@ public class UpdateMappingTests extends ElasticsearchIntegrationTest {
@SuppressWarnings("unchecked")
@Test
public void updateIncludeExclude() throws Exception {
assertAcked(prepareCreate("test").addMapping("type",
assertAcked(prepareCreate("test").addMapping("type",
jsonBuilder().startObject().startObject("type").startObject("properties")
.startObject("normal").field("type", "long").endObject()
.startObject("exclude").field("type", "long").endObject()
.startObject("include").field("type", "long").endObject()
.endObject().endObject().endObject()));
.startObject("normal").field("type", "long").endObject()
.startObject("exclude").field("type", "long").endObject()
.startObject("include").field("type", "long").endObject()
.endObject().endObject().endObject()));
ensureGreen(); // make sure that replicas are initialized so the refresh command will work them too
logger.info("Index doc");