[GitHub] [incubator-druid] jon-wei commented on a change in pull request #8287: Cleanup appenderators and segment walkers in UnifiedIndexerAppenderatorsManager

2019-08-20 Thread GitBox
jon-wei commented on a change in pull request #8287: Cleanup appenderators and 
segment walkers in UnifiedIndexerAppenderatorsManager
URL: https://github.com/apache/incubator-druid/pull/8287#discussion_r315943072
 
 

 ##
 File path: 
server/src/main/java/org/apache/druid/segment/realtime/appenderator/UnifiedIndexerAppenderatorsManager.java
 ##
 @@ -183,22 +192,48 @@ public Appenderator createOfflineAppenderatorForTask(
   IndexMerger indexMerger
   )
   {
-Appenderator appenderator = Appenderators.createOffline(
-schema,
-rewriteAppenderatorConfigMemoryLimits(config),
-metrics,
-dataSegmentPusher,
-objectMapper,
-indexIO,
-wrapIndexMerger(indexMerger)
-);
-return appenderator;
+synchronized (this) {
+  DatasourceBundle datasourceBundle = datasourceBundles.computeIfAbsent(
+  schema.getDataSource(),
+  (datasource) -> {
+return new DatasourceBundle(datasource);
+  }
+  );
+
+  Appenderator appenderator = Appenderators.createOffline(
+  schema,
+  rewriteAppenderatorConfigMemoryLimits(config),
+  metrics,
+  dataSegmentPusher,
+  objectMapper,
+  indexIO,
+  wrapIndexMerger(indexMerger)
+  );
+  datasourceBundle.addAppenderator(taskId, appenderator);
+  return appenderator;
+}
   }
 
   @Override
-  public void removeAppenderatorForTask(String taskId)
+  public void removeAppenderatorsForTask(
+  String taskId,
+  String dataSource
+  )
   {
-// nothing to remove presently
+synchronized (this) {
+  DatasourceBundle datasourceBundle = datasourceBundles.get(dataSource);
+  if (datasourceBundle == null) {
+LOG.error("Could not find datasource bundle for [%s], task [%s]", 
dataSource, taskId);
 
 Review comment:
   Hm, I think it's fine as ERROR for now since the operator can't really take 
any action to remedy the situation, and I don't think the error is severe 
enough to warrant alerting. We could revisit this area in the future if needed.


This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

-
To unsubscribe, e-mail: commits-unsubscr...@druid.apache.org
For additional commands, e-mail: commits-h...@druid.apache.org



[GitHub] [incubator-druid] jon-wei commented on a change in pull request #8287: Cleanup appenderators and segment walkers in UnifiedIndexerAppenderatorsManager

2019-08-15 Thread GitBox
jon-wei commented on a change in pull request #8287: Cleanup appenderators and 
segment walkers in UnifiedIndexerAppenderatorsManager
URL: https://github.com/apache/incubator-druid/pull/8287#discussion_r314547893
 
 

 ##
 File path: 
server/src/main/java/org/apache/druid/segment/realtime/appenderator/UnifiedIndexerAppenderatorsManager.java
 ##
 @@ -248,6 +289,50 @@ private AppenderatorConfig 
rewriteAppenderatorConfigMemoryLimits(AppenderatorCon
 return new MemoryParameterOverridingAppenderatorConfig(baseConfig, 
perWorkerLimit);
   }
 
+  private class DatasourceBundle
+  {
+private final SinkQuerySegmentWalker walker;
+private final Map> taskAppenderatorMap;
 
 Review comment:
   The CompactionTask creates multiple IndexTask instances and runs them in 
sequence (each will create its own appenderator). I updated the CompactionTask 
to use the same task ID for the index tasks it runs internally.


This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

-
To unsubscribe, e-mail: commits-unsubscr...@druid.apache.org
For additional commands, e-mail: commits-h...@druid.apache.org



[GitHub] [incubator-druid] jon-wei commented on a change in pull request #8287: Cleanup appenderators and segment walkers in UnifiedIndexerAppenderatorsManager

2019-08-15 Thread GitBox
jon-wei commented on a change in pull request #8287: Cleanup appenderators and 
segment walkers in UnifiedIndexerAppenderatorsManager
URL: https://github.com/apache/incubator-druid/pull/8287#discussion_r314547644
 
 

 ##
 File path: 
server/src/main/java/org/apache/druid/segment/realtime/appenderator/UnifiedIndexerAppenderatorsManager.java
 ##
 @@ -248,6 +289,50 @@ private AppenderatorConfig 
rewriteAppenderatorConfigMemoryLimits(AppenderatorCon
 return new MemoryParameterOverridingAppenderatorConfig(baseConfig, 
perWorkerLimit);
   }
 
+  private class DatasourceBundle
+  {
+private final SinkQuerySegmentWalker walker;
+private final Map> taskAppenderatorMap;
+
+public DatasourceBundle(
+String dataSource
+)
+{
+  this.taskAppenderatorMap = new HashMap<>();
+
+  VersionedIntervalTimeline sinkTimeline = new 
VersionedIntervalTimeline<>(
+  String.CASE_INSENSITIVE_ORDER
+  );
+  this.walker = new SinkQuerySegmentWalker(
+  dataSource,
+  sinkTimeline,
+  objectMapper,
+  serviceEmitter,
+  queryRunnerFactoryConglomerateProvider.get(),
+  queryExecutorService,
+  Preconditions.checkNotNull(cache, "cache"),
+  cacheConfig,
+  cachePopulatorStats
+  );
+}
+
+public SinkQuerySegmentWalker getWalker()
+{
+  return walker;
+}
+
+public void addAppenderator(String taskId, Appenderator appenderator)
+{
+  taskAppenderatorMap.computeIfAbsent(
+  taskId,
+  (myTaskId) -> {
+return new ArrayList<>();
+  }
+  );
+  taskAppenderatorMap.get(taskId).add(appenderator);
 
 Review comment:
   Changed, thanks


This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

-
To unsubscribe, e-mail: commits-unsubscr...@druid.apache.org
For additional commands, e-mail: commits-h...@druid.apache.org



[GitHub] [incubator-druid] jon-wei commented on a change in pull request #8287: Cleanup appenderators and segment walkers in UnifiedIndexerAppenderatorsManager

2019-08-15 Thread GitBox
jon-wei commented on a change in pull request #8287: Cleanup appenderators and 
segment walkers in UnifiedIndexerAppenderatorsManager
URL: https://github.com/apache/incubator-druid/pull/8287#discussion_r314547560
 
 

 ##
 File path: 
server/src/main/java/org/apache/druid/segment/realtime/appenderator/UnifiedIndexerAppenderatorsManager.java
 ##
 @@ -183,22 +192,48 @@ public Appenderator createOfflineAppenderatorForTask(
   IndexMerger indexMerger
   )
   {
-Appenderator appenderator = Appenderators.createOffline(
-schema,
-rewriteAppenderatorConfigMemoryLimits(config),
-metrics,
-dataSegmentPusher,
-objectMapper,
-indexIO,
-wrapIndexMerger(indexMerger)
-);
-return appenderator;
+synchronized (this) {
+  DatasourceBundle datasourceBundle = datasourceBundles.computeIfAbsent(
+  schema.getDataSource(),
+  (datasource) -> {
+return new DatasourceBundle(datasource);
+  }
+  );
+
+  Appenderator appenderator = Appenderators.createOffline(
+  schema,
+  rewriteAppenderatorConfigMemoryLimits(config),
+  metrics,
+  dataSegmentPusher,
+  objectMapper,
+  indexIO,
+  wrapIndexMerger(indexMerger)
+  );
+  datasourceBundle.addAppenderator(taskId, appenderator);
+  return appenderator;
+}
   }
 
   @Override
-  public void removeAppenderatorForTask(String taskId)
+  public void removeAppenderatorsForTask(
+  String taskId,
+  String dataSource
+  )
   {
-// nothing to remove presently
+synchronized (this) {
+  DatasourceBundle datasourceBundle = datasourceBundles.get(dataSource);
+  if (datasourceBundle == null) {
+LOG.error("Could not find datasource bundle for [%s], task [%s]", 
dataSource, taskId);
 
 Review comment:
   I think this is more of an error at the level of the task manager and it 
would only be called after a task has finished (so I don't think it could be 
propagated to tasks), it's not fatal enough to stop altogether but I think it 
indicates bugs so I wanted to be ERROR instead of WARN. 
   
   
   


This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

-
To unsubscribe, e-mail: commits-unsubscr...@druid.apache.org
For additional commands, e-mail: commits-h...@druid.apache.org



[GitHub] [incubator-druid] jon-wei commented on a change in pull request #8287: Cleanup appenderators and segment walkers in UnifiedIndexerAppenderatorsManager

2019-08-15 Thread GitBox
jon-wei commented on a change in pull request #8287: Cleanup appenderators and 
segment walkers in UnifiedIndexerAppenderatorsManager
URL: https://github.com/apache/incubator-druid/pull/8287#discussion_r314546982
 
 

 ##
 File path: 
server/src/main/java/org/apache/druid/segment/realtime/appenderator/UnifiedIndexerAppenderatorsManager.java
 ##
 @@ -183,22 +192,48 @@ public Appenderator createOfflineAppenderatorForTask(
   IndexMerger indexMerger
   )
   {
-Appenderator appenderator = Appenderators.createOffline(
-schema,
-rewriteAppenderatorConfigMemoryLimits(config),
-metrics,
-dataSegmentPusher,
-objectMapper,
-indexIO,
-wrapIndexMerger(indexMerger)
-);
-return appenderator;
+synchronized (this) {
 
 Review comment:
   I decided to leave it as is, since the shared portion is pretty small here 
and I feel it's more readable without the additional indirection


This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

-
To unsubscribe, e-mail: commits-unsubscr...@druid.apache.org
For additional commands, e-mail: commits-h...@druid.apache.org



[GitHub] [incubator-druid] jon-wei commented on a change in pull request #8287: Cleanup appenderators and segment walkers in UnifiedIndexerAppenderatorsManager

2019-08-15 Thread GitBox
jon-wei commented on a change in pull request #8287: Cleanup appenderators and 
segment walkers in UnifiedIndexerAppenderatorsManager
URL: https://github.com/apache/incubator-druid/pull/8287#discussion_r314546755
 
 

 ##
 File path: 
server/src/main/java/org/apache/druid/segment/realtime/appenderator/UnifiedIndexerAppenderatorsManager.java
 ##
 @@ -134,41 +154,30 @@ public Appenderator createRealtimeAppenderatorForTask(
   CachePopulatorStats cachePopulatorStats
   )
   {
-SinkQuerySegmentWalker segmentWalker = 
datasourceSegmentWalkers.computeIfAbsent(
-schema.getDataSource(),
-(datasource) -> {
-  VersionedIntervalTimeline sinkTimeline = new 
VersionedIntervalTimeline<>(
-  String.CASE_INSENSITIVE_ORDER
-  );
-  SinkQuerySegmentWalker datasourceSegmentWalker = new 
SinkQuerySegmentWalker(
-  schema.getDataSource(),
-  sinkTimeline,
-  objectMapper,
-  emitter,
-  conglomerate,
-  this.queryExecutorService,
-  Preconditions.checkNotNull(this.cache, "cache"),
-  this.cacheConfig,
-  this.cachePopulatorStats
-  );
-  return datasourceSegmentWalker;
-}
-);
+synchronized (this) {
+  DatasourceBundle datasourceBundle = datasourceBundles.computeIfAbsent(
+  schema.getDataSource(),
+  (datasource) -> {
+return new DatasourceBundle(datasource);
 
 Review comment:
   Changed to `DatasourceBundle::new`


This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

-
To unsubscribe, e-mail: commits-unsubscr...@druid.apache.org
For additional commands, e-mail: commits-h...@druid.apache.org



[GitHub] [incubator-druid] jon-wei commented on a change in pull request #8287: Cleanup appenderators and segment walkers in UnifiedIndexerAppenderatorsManager

2019-08-15 Thread GitBox
jon-wei commented on a change in pull request #8287: Cleanup appenderators and 
segment walkers in UnifiedIndexerAppenderatorsManager
URL: https://github.com/apache/incubator-druid/pull/8287#discussion_r314546703
 
 

 ##
 File path: 
server/src/main/java/org/apache/druid/segment/realtime/appenderator/UnifiedIndexerAppenderatorsManager.java
 ##
 @@ -248,6 +289,50 @@ private AppenderatorConfig 
rewriteAppenderatorConfigMemoryLimits(AppenderatorCon
 return new MemoryParameterOverridingAppenderatorConfig(baseConfig, 
perWorkerLimit);
   }
 
+  private class DatasourceBundle
+  {
+private final SinkQuerySegmentWalker walker;
+private final Map> taskAppenderatorMap;
+
+public DatasourceBundle(
 
 Review comment:
   Hm, I made the class public and added a `@VisibleForTesting` annotation, 
since it's accessed from `TaskLifecycleTest` now


This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


With regards,
Apache Git Services

-
To unsubscribe, e-mail: commits-unsubscr...@druid.apache.org
For additional commands, e-mail: commits-h...@druid.apache.org