org.apache.druid.server

Class QueryScheduler

java.lang.Object

org.apache.druid.server.QueryScheduler

All Implemented Interfaces:

QueryWatcher

public class QueryScheduler
extends Object
implements QueryWatcher

QueryScheduler (potentially) assigns any Query that is to be executed to a 'query lane' using the QueryLaningStrategy that is defined in QuerySchedulerConfig. As a QueryWatcher, it also provides cancellation facilities to brokers, historicals, and realtime tasks. This class is shared by all requests on the HTTP theadpool and must be thread safe.

Field Summary

Fields

Modifier and Type	Field and Description
static String	TOTAL
static int	UNAVAILABLE

Constructor Summary

Constructors

Constructor and Description
QueryScheduler(int totalNumThreads, QueryPrioritizationStrategy prioritizationStrategy, QueryLaningStrategy laningStrategy, ServerConfig serverConfig) Keeping the old constructor as many test classes are dependent on this
QueryScheduler(int totalNumThreads, QueryPrioritizationStrategy prioritizationStrategy, QueryLaningStrategy laningStrategy, ServerConfig serverConfig, ServiceEmitter emitter)

Method Summary

Modifier and Type	Method and Description
boolean	cancelQuery(String id) Forcibly cancel all futures that have been registered to a specific query id
Set<String>	getQueryDatasources(String queryId) Get a Set of datasource names for a Query id, used by QueryResource.cancelQuery(java.lang.String, javax.servlet.http.HttpServletRequest) to authorize that a user may call cancelQuery(java.lang.String) for the given id and datasources
<T> Query<T>	prioritizeAndLaneQuery(QueryPlus<T> queryPlus, Set<SegmentServerSelector> segments) Assign a query a priority and lane (if not set)
void	registerQueryFuture(Query<?> query, com.google.common.util.concurrent.ListenableFuture<?> future)
<T> Sequence<T>	run(Query<?> query, Sequence<T> resultSequence) Run a query with the scheduler, attempting to acquire a semaphore from the total and lane specific query capacities Note that cancelQuery(java.lang.String) should not interrupt the thread that calls run, in all current usages it only cancels any ListenableFuture created downstream.
<T> QueryRunner<T>	wrapQueryRunner(QueryRunner<T> baseRunner) Returns a QueryRunner that will call run(org.apache.druid.query.Query<?>, org.apache.druid.java.util.common.guava.Sequence<T>) when QueryRunner.run(org.apache.druid.query.QueryPlus<T>, org.apache.druid.query.context.ResponseContext) is called.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

UNAVAILABLE

public static final int UNAVAILABLE

See Also:

Constant Field Values

TOTAL

public static final String TOTAL

See Also:

Constant Field Values

Constructor Detail

QueryScheduler

public QueryScheduler(int totalNumThreads,
                      QueryPrioritizationStrategy prioritizationStrategy,
                      QueryLaningStrategy laningStrategy,
                      ServerConfig serverConfig,
                      ServiceEmitter emitter)

QueryScheduler

public QueryScheduler(int totalNumThreads,
                      QueryPrioritizationStrategy prioritizationStrategy,
                      QueryLaningStrategy laningStrategy,
                      ServerConfig serverConfig)

Keeping the old constructor as many test classes are dependent on this

Method Detail

registerQueryFuture

public void registerQueryFuture(Query<?> query,
                                com.google.common.util.concurrent.ListenableFuture<?> future)

Specified by:

registerQueryFuture in interface QueryWatcher

prioritizeAndLaneQuery

public <T> Query<T> prioritizeAndLaneQuery(QueryPlus<T> queryPlus,
                                           Set<SegmentServerSelector> segments)

Assign a query a priority and lane (if not set)

run

public <T> Sequence<T> run(Query<?> query,
                           Sequence<T> resultSequence)

Run a query with the scheduler, attempting to acquire a semaphore from the total and lane specific query capacities Note that cancelQuery(java.lang.String) should not interrupt the thread that calls run, in all current usages it only cancels any ListenableFuture created downstream. If this ever commonly changes, we should add synchronization between cancelQuery(java.lang.String) and the acquisition of the Bulkhead to continue to ensure that anything acquired is also released. In the meantime, if a ListenableFuture is registered for the query that calls this method, it MUST handle this synchronization itself to ensure that no Bulkhead is acquired without releasing it.

wrapQueryRunner

public <T> QueryRunner<T> wrapQueryRunner(QueryRunner<T> baseRunner)

Returns a QueryRunner that will call run(org.apache.druid.query.Query<?>, org.apache.druid.java.util.common.guava.Sequence<T>) when QueryRunner.run(org.apache.druid.query.QueryPlus<T>, org.apache.druid.query.context.ResponseContext) is called.

cancelQuery

public boolean cancelQuery(String id)

Forcibly cancel all futures that have been registered to a specific query id

getQueryDatasources

public Set<String> getQueryDatasources(String queryId)

Get a Set of datasource names for a Query id, used by QueryResource.cancelQuery(java.lang.String, javax.servlet.http.HttpServletRequest) to authorize that a user may call cancelQuery(java.lang.String) for the given id and datasources