Reduce default max # of subTasks to 1 for native parallel task (#7181)

* Reduce # of max subTasks to 2

* fix typo and add more doc

* add more doc and link

* change default and add warning

* fix doc

* add test

* fix it test
This commit is contained in:
Jihoon Son 2019-03-05 22:06:36 -08:00 committed by Fangjin Yang
parent 9183e32876
commit e48a9c138e
8 changed files with 162 additions and 78 deletions

View File

@ -54,7 +54,17 @@ which specifies a split and submits worker tasks using those specs. As a result,
the implementation of splittable firehoses. Please note that multiple tasks can be created for the same worker task spec
if one of them fails.
Since this task doesn't shuffle intermediate data, it isn't available for [perfect rollup](../ingestion/index.html#roll-up-modes).
You may want to consider the below points:
- Since this task doesn't shuffle intermediate data, it isn't available for [perfect rollup](../ingestion/index.html#roll-up-modes).
- The number of tasks for parallel ingestion is decided by `maxNumSubTasks` in the tuningConfig.
Since the supervisor task creates up to `maxNumSubTasks` worker tasks regardless of the available task slots,
it may affect to other ingestion performance. As a result, it's important to set `maxNumSubTasks` properly.
See the below [Capacity Planning](#capacity-planning) section for more details.
- By default, batch ingestion replaces all data in any segment that it writes to. If you'd like to add to the segment
instead, set the appendToExisting flag in ioConfig. Note that it only replaces data in segments where it actively adds
data: if there are segments in your granularitySpec's intervals that have no data written by this task, they will be
left alone.
An example ingestion spec is:
@ -122,16 +132,15 @@ An example ingestion spec is:
"baseDir": "examples/indexing/",
"filter": "wikipedia_index_data*"
}
},
"tuningconfig": {
"type": "index_parallel",
"maxNumSubTasks": 2
}
}
}
```
By default, batch ingestion replaces all data in any segment that it writes to. If you'd like to add to the segment
instead, set the appendToExisting flag in ioConfig. Note that it only replaces data in segments where it actively adds
data: if there are segments in your granularitySpec's intervals that have no data written by this task, they will be
left alone.
#### Task Properties
|property|description|required?|
@ -181,7 +190,7 @@ The tuningConfig is optional and default parameters will be used if no tuningCon
|reportParseExceptions|If true, exceptions encountered during parsing will be thrown and will halt ingestion; if false, unparseable rows and fields will be skipped.|false|no|
|pushTimeout|Milliseconds to wait for pushing segments. It must be >= 0, where 0 means to wait forever.|0|no|
|segmentWriteOutMediumFactory|Segment write-out medium to use when creating segments. See [SegmentWriteOutMediumFactory](#segmentWriteOutMediumFactory).|Not specified, the value from `druid.peon.defaultSegmentWriteOutMediumFactory.type` is used|no|
|maxNumSubTasks|Maximum number of tasks which can be run at the same time.|Integer.MAX_VALUE|no|
|maxNumSubTasks|Maximum number of tasks which can be run at the same time. The supervisor task would spawn worker tasks up to `maxNumSubTasks` regardless of the available task slots. If this value is set to 1, the supervisor task processes data ingestion on its own instead of spawning worker tasks. If this value is set to too large, too many worker tasks can be created which might block other ingestion. Check [Capacity Planning](#capacity-planning) for more details.|1|no|
|maxRetry|Maximum number of retries on task failures.|3|no|
|taskStatusCheckPeriodMs|Polling period in milleseconds to check running task statuses.|1000|no|
|chatHandlerTimeout|Timeout for reporting the pushed segments in worker tasks.|PT10S|no|
@ -372,7 +381,7 @@ An example of the result is
"reportParseExceptions": false,
"pushTimeout": 0,
"segmentWriteOutMediumFactory": null,
"maxNumSubTasks": 2147483647,
"maxNumSubTasks": 4,
"maxRetry": 3,
"taskStatusCheckPeriodMs": 1000,
"chatHandlerTimeout": "PT10S",
@ -408,6 +417,27 @@ An example of the result is
Returns the task attempt history of the worker task spec of the given id, or HTTP 404 Not Found error if the supervisor task is running in the sequential mode.
### Capacity Planning
The supervisor task can create up to `maxNumSubTasks` worker tasks no matter how many task slots are currently available.
As a result, total number of tasks which can be run at the same time is `(maxNumSubTasks + 1)` (including the supervisor task).
Please note that this can be even larger than total number of task slots (sum of the capacity of all workers).
If `maxNumSubTasks` is larger than `n (available task slots)`, then
`maxNumSubTasks` tasks are created by the supervisor task, but only `n` tasks would be started.
Others will wait in the pending state until any running task is finished.
If you are using the Parallel Index Task with stream ingestion together,
we would recommend to limit the max capacity for batch ingestion to prevent
stream ingestion from being blocked by batch ingestion. Suppose you have
`t` Parallel Index Tasks to run at the same time, but want to limit
the max number of tasks for batch ingestion to `b`. Then, (sum of `maxNumSubTasks`
of all Parallel Index Tasks + `t` (for supervisor tasks)) must be smaller than `b`.
If you have some tasks of a higher priority than others, you may set their
`maxNumSubTasks` to a higher value than lower priority tasks.
This may help the higher priority tasks to finish earlier than lower priority tasks
by assigning more task slots to them.
Local Index Task
----------------

View File

@ -405,7 +405,16 @@ public class IndexTask extends AbstractTask implements ChatHandler
try {
if (chatHandlerProvider.isPresent()) {
log.info("Found chat handler of class[%s]", chatHandlerProvider.get().getClass().getName());
if (chatHandlerProvider.get().get(getId()).isPresent()) {
// This is a workaround for ParallelIndexSupervisorTask to avoid double registering when it runs in the
// sequential mode. See ParallelIndexSupervisorTask.runSequential().
// Note that all HTTP endpoints are not available in this case. This works only for
// ParallelIndexSupervisorTask because it doesn't support APIs for live ingestion reports.
log.warn("Chat handler is already registered. Skipping chat handler registration.");
} else {
chatHandlerProvider.get().register(getId(), this, false);
}
} else {
log.warn("No chat handler detected");
}

View File

@ -256,13 +256,23 @@ public class ParallelIndexSupervisorTask extends AbstractTask implements ChatHan
chatHandlerProvider.register(getId(), this, false);
try {
if (baseFirehoseFactory.isSplittable()) {
if (isParallelMode()) {
return runParallel(toolbox);
} else {
if (!baseFirehoseFactory.isSplittable()) {
log.warn(
"firehoseFactory[%s] is not splittable. Running sequentially",
"firehoseFactory[%s] is not splittable. Running sequentially.",
baseFirehoseFactory.getClass().getSimpleName()
);
} else if (ingestionSchema.getTuningConfig().getMaxNumSubTasks() == 1) {
log.warn(
"maxNumSubTasks is 1. Running sequentially. "
+ "Please set maxNumSubTasks to something higher than 1 if you want to run in parallel ingestion mode."
);
} else {
throw new ISE("Unknown reason for sequentail mode. Failing this task.");
}
return runSequential(toolbox);
}
}
@ -271,6 +281,15 @@ public class ParallelIndexSupervisorTask extends AbstractTask implements ChatHan
}
}
private boolean isParallelMode()
{
if (baseFirehoseFactory.isSplittable() && ingestionSchema.getTuningConfig().getMaxNumSubTasks() > 1) {
return true;
} else {
return false;
}
}
@VisibleForTesting
void setToolbox(TaskToolbox toolbox)
{
@ -280,7 +299,7 @@ public class ParallelIndexSupervisorTask extends AbstractTask implements ChatHan
private TaskStatus runParallel(TaskToolbox toolbox) throws Exception
{
createRunner(toolbox);
return TaskStatus.fromCode(getId(), runner.run());
return TaskStatus.fromCode(getId(), Preconditions.checkNotNull(runner, "runner").run());
}
private TaskStatus runSequential(TaskToolbox toolbox)
@ -479,11 +498,7 @@ public class ParallelIndexSupervisorTask extends AbstractTask implements ChatHan
public Response getMode(@Context final HttpServletRequest req)
{
IndexTaskUtils.datasourceAuthorizationCheck(req, Action.READ, getDataSource(), authorizerMapper);
if (runner == null) {
return Response.status(Response.Status.SERVICE_UNAVAILABLE).entity("task is not running yet").build();
} else {
return Response.ok(baseFirehoseFactory.isSplittable() ? "parallel" : "sequential").build();
}
return Response.ok(isParallelMode() ? "parallel" : "sequential").build();
}
@GET

View File

@ -22,6 +22,7 @@ package org.apache.druid.indexing.common.task.batch.parallel;
import com.fasterxml.jackson.annotation.JsonCreator;
import com.fasterxml.jackson.annotation.JsonProperty;
import com.fasterxml.jackson.annotation.JsonTypeName;
import com.google.common.base.Preconditions;
import org.apache.druid.indexing.common.task.IndexTask.IndexTuningConfig;
import org.apache.druid.segment.IndexSpec;
import org.apache.druid.segment.writeout.SegmentWriteOutMediumFactory;
@ -34,7 +35,7 @@ import java.util.Objects;
@JsonTypeName("index_parallel")
public class ParallelIndexTuningConfig extends IndexTuningConfig
{
private static final int DEFAULT_MAX_NUM_BATCH_TASKS = Integer.MAX_VALUE; // unlimited
private static final int DEFAULT_MAX_NUM_BATCH_TASKS = 1;
private static final int DEFAULT_MAX_RETRY = 3;
private static final long DEFAULT_TASK_STATUS_CHECK_PERIOD_MS = 1000;
@ -131,6 +132,8 @@ public class ParallelIndexTuningConfig extends IndexTuningConfig
this.chatHandlerTimeout = DEFAULT_CHAT_HANDLER_TIMEOUT;
this.chatHandlerNumRetries = DEFAULT_CHAT_HANDLER_NUM_RETRIES;
Preconditions.checkArgument(this.maxNumSubTasks > 0, "maxNumSubTasks must be positive");
}
@JsonProperty

View File

@ -294,22 +294,6 @@ public class AbstractParallelIndexSupervisorTaskTest extends IngestionTestBase
new DropwizardRowIngestionMetersFactory()
);
}
@Override
public TaskStatus run(TaskToolbox toolbox) throws Exception
{
return TaskStatus.fromCode(
getId(),
new TestParallelIndexTaskRunner(
toolbox,
getId(),
getGroupId(),
getIngestionSchema(),
getContext(),
new NoopIndexingServiceClient()
).run()
);
}
}
static class TestParallelIndexTaskRunner extends SinglePhaseParallelIndexTaskRunner

View File

@ -281,9 +281,8 @@ public class ParallelIndexSupervisorTaskKillTest extends AbstractParallelIndexSu
}
@Override
public TaskStatus run(TaskToolbox toolbox) throws Exception
ParallelIndexTaskRunner createRunner(TaskToolbox toolbox)
{
setToolbox(toolbox);
setRunner(
new TestRunner(
toolbox,
@ -291,10 +290,7 @@ public class ParallelIndexSupervisorTaskKillTest extends AbstractParallelIndexSu
indexingServiceClient
)
);
return TaskStatus.fromCode(
getId(),
getRunner().run()
);
return getRunner();
}
}

View File

@ -500,7 +500,7 @@ public class ParallelIndexSupervisorTaskResourceTest extends AbstractParallelInd
}
@Override
public TaskStatus run(TaskToolbox toolbox) throws Exception
ParallelIndexTaskRunner createRunner(TaskToolbox toolbox)
{
setRunner(
new TestRunner(
@ -509,10 +509,7 @@ public class ParallelIndexSupervisorTaskResourceTest extends AbstractParallelInd
indexingServiceClient
)
);
return TaskStatus.fromCode(
getId(),
getRunner().run()
);
return getRunner();
}
}

View File

@ -24,7 +24,6 @@ import org.apache.druid.data.input.FiniteFirehoseFactory;
import org.apache.druid.data.input.InputSplit;
import org.apache.druid.data.input.impl.StringInputRowParser;
import org.apache.druid.indexer.TaskState;
import org.apache.druid.indexer.TaskStatus;
import org.apache.druid.indexing.common.TaskToolbox;
import org.apache.druid.indexing.common.actions.TaskActionClient;
import org.apache.druid.indexing.common.task.TaskResource;
@ -229,10 +228,87 @@ public class ParallelIndexSupervisorTaskTest extends AbstractParallelIndexSuperv
Assert.assertEquals(TaskState.SUCCESS, task.run(toolbox).getStatusCode());
}
@Test
public void testWith1MaxNumSubTasks() throws Exception
{
final ParallelIndexSupervisorTask task = newTask(
Intervals.of("2017/2018"),
new ParallelIndexIOConfig(
new LocalFirehoseFactory(inputDir, "test_*", null),
false
),
new ParallelIndexTuningConfig(
null,
null,
null,
null,
null,
null,
null,
null,
null,
null,
null,
null,
null,
1,
null,
null,
null,
null,
null,
null,
null
)
);
actionClient = createActionClient(task);
toolbox = createTaskToolbox(task);
prepareTaskForLocking(task);
Assert.assertTrue(task.isReady(actionClient));
Assert.assertEquals(TaskState.SUCCESS, task.run(toolbox).getStatusCode());
Assert.assertNull("Runner must be null if the task was in the sequential mode", task.getRunner());
}
private ParallelIndexSupervisorTask newTask(
Interval interval,
ParallelIndexIOConfig ioConfig
)
{
return newTask(
interval,
ioConfig,
new ParallelIndexTuningConfig(
null,
null,
null,
null,
null,
null,
null,
null,
null,
null,
null,
null,
null,
2,
null,
null,
null,
null,
null,
null,
null
)
);
}
private ParallelIndexSupervisorTask newTask(
Interval interval,
ParallelIndexIOConfig ioConfig,
ParallelIndexTuningConfig tuningConfig
)
{
// set up ingestion spec
final ParallelIndexIngestionSpec ingestionSpec = new ParallelIndexIngestionSpec(
@ -257,29 +333,7 @@ public class ParallelIndexSupervisorTaskTest extends AbstractParallelIndexSuperv
getObjectMapper()
),
ioConfig,
new ParallelIndexTuningConfig(
null,
null,
null,
null,
null,
null,
null,
null,
null,
null,
null,
null,
null,
2,
null,
null,
null,
null,
null,
null,
null
)
tuningConfig
);
// set up test tools
@ -315,9 +369,8 @@ public class ParallelIndexSupervisorTaskTest extends AbstractParallelIndexSuperv
}
@Override
public TaskStatus run(TaskToolbox toolbox) throws Exception
ParallelIndexTaskRunner createRunner(TaskToolbox toolbox)
{
setToolbox(toolbox);
setRunner(
new TestRunner(
toolbox,
@ -325,10 +378,7 @@ public class ParallelIndexSupervisorTaskTest extends AbstractParallelIndexSuperv
indexingServiceClient
)
);
return TaskStatus.fromCode(
getId(),
getRunner().run()
);
return getRunner();
}
}