HLRC: ML start data feed API (#33898)

* HLRC: ML start data feed API
This commit is contained in:
Benjamin Trent 2018-09-21 05:56:38 -07:00 committed by GitHub
parent 1de2a925ce
commit bf0a0f74da
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
11 changed files with 647 additions and 0 deletions

View File

@ -48,6 +48,7 @@ import org.elasticsearch.client.ml.PostDataRequest;
import org.elasticsearch.client.ml.PutCalendarRequest; import org.elasticsearch.client.ml.PutCalendarRequest;
import org.elasticsearch.client.ml.PutDatafeedRequest; import org.elasticsearch.client.ml.PutDatafeedRequest;
import org.elasticsearch.client.ml.PutJobRequest; import org.elasticsearch.client.ml.PutJobRequest;
import org.elasticsearch.client.ml.StartDatafeedRequest;
import org.elasticsearch.client.ml.UpdateJobRequest; import org.elasticsearch.client.ml.UpdateJobRequest;
import org.elasticsearch.common.Strings; import org.elasticsearch.common.Strings;
import org.elasticsearch.common.bytes.BytesReference; import org.elasticsearch.common.bytes.BytesReference;
@ -231,6 +232,19 @@ final class MLRequestConverters {
return request; return request;
} }
static Request startDatafeed(StartDatafeedRequest startDatafeedRequest) throws IOException {
String endpoint = new EndpointBuilder()
.addPathPartAsIs("_xpack")
.addPathPartAsIs("ml")
.addPathPartAsIs("datafeeds")
.addPathPart(startDatafeedRequest.getDatafeedId())
.addPathPartAsIs("_start")
.build();
Request request = new Request(HttpPost.METHOD_NAME, endpoint);
request.setEntity(createEntity(startDatafeedRequest, REQUEST_BODY_CONTENT_TYPE));
return request;
}
static Request deleteForecast(DeleteForecastRequest deleteForecastRequest) { static Request deleteForecast(DeleteForecastRequest deleteForecastRequest) {
String endpoint = new EndpointBuilder() String endpoint = new EndpointBuilder()
.addPathPartAsIs("_xpack") .addPathPartAsIs("_xpack")

View File

@ -58,6 +58,8 @@ import org.elasticsearch.client.ml.PutDatafeedRequest;
import org.elasticsearch.client.ml.PutDatafeedResponse; import org.elasticsearch.client.ml.PutDatafeedResponse;
import org.elasticsearch.client.ml.PutJobRequest; import org.elasticsearch.client.ml.PutJobRequest;
import org.elasticsearch.client.ml.PutJobResponse; import org.elasticsearch.client.ml.PutJobResponse;
import org.elasticsearch.client.ml.StartDatafeedRequest;
import org.elasticsearch.client.ml.StartDatafeedResponse;
import org.elasticsearch.client.ml.UpdateJobRequest; import org.elasticsearch.client.ml.UpdateJobRequest;
import org.elasticsearch.client.ml.job.stats.JobStats; import org.elasticsearch.client.ml.job.stats.JobStats;
@ -565,6 +567,46 @@ public final class MachineLearningClient {
Collections.emptySet()); Collections.emptySet());
} }
/**
* Starts the given Machine Learning Datafeed
* <p>
* For additional info
* see <a href="http://www.elastic.co/guide/en/elasticsearch/reference/current/ml-start-datafeed.html">
* ML Start Datafeed documentation</a>
*
* @param request The request to start the datafeed
* @param options Additional request options (e.g. headers), use {@link RequestOptions#DEFAULT} if nothing needs to be customized
* @return action acknowledgement
* @throws IOException when there is a serialization issue sending the request or receiving the response
*/
public StartDatafeedResponse startDatafeed(StartDatafeedRequest request, RequestOptions options) throws IOException {
return restHighLevelClient.performRequestAndParseEntity(request,
MLRequestConverters::startDatafeed,
options,
StartDatafeedResponse::fromXContent,
Collections.emptySet());
}
/**
* Starts the given Machine Learning Datafeed asynchronously and notifies the listener on completion
* <p>
* For additional info
* see <a href="http://www.elastic.co/guide/en/elasticsearch/reference/current/ml-start-datafeed.html">
* ML Start Datafeed documentation</a>
*
* @param request The request to start the datafeed
* @param options Additional request options (e.g. headers), use {@link RequestOptions#DEFAULT} if nothing needs to be customized
* @param listener Listener to be notified upon request completion
*/
public void startDatafeedAsync(StartDatafeedRequest request, RequestOptions options, ActionListener<StartDatafeedResponse> listener) {
restHighLevelClient.performRequestAsyncAndParseEntity(request,
MLRequestConverters::startDatafeed,
options,
StartDatafeedResponse::fromXContent,
listener,
Collections.emptySet());
}
/** /**
* Updates a Machine Learning {@link org.elasticsearch.client.ml.job.config.Job} * Updates a Machine Learning {@link org.elasticsearch.client.ml.job.config.Job}
* <p> * <p>

View File

@ -0,0 +1,160 @@
/*
* Licensed to Elasticsearch under one or more contributor
* license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright
* ownership. Elasticsearch licenses this file to you under
* the Apache License, Version 2.0 (the "License"); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing,
* software distributed under the License is distributed on an
* "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
* KIND, either express or implied. See the License for the
* specific language governing permissions and limitations
* under the License.
*/
package org.elasticsearch.client.ml;
import org.elasticsearch.action.ActionRequest;
import org.elasticsearch.action.ActionRequestValidationException;
import org.elasticsearch.client.ml.datafeed.DatafeedConfig;
import org.elasticsearch.common.ParseField;
import org.elasticsearch.common.unit.TimeValue;
import org.elasticsearch.common.xcontent.ConstructingObjectParser;
import org.elasticsearch.common.xcontent.ToXContentObject;
import org.elasticsearch.common.xcontent.XContentBuilder;
import java.io.IOException;
import java.util.Objects;
/**
* Request to start a Datafeed
*/
public class StartDatafeedRequest extends ActionRequest implements ToXContentObject {
public static final ParseField START = new ParseField("start");
public static final ParseField END = new ParseField("end");
public static final ParseField TIMEOUT = new ParseField("timeout");
public static ConstructingObjectParser<StartDatafeedRequest, Void> PARSER =
new ConstructingObjectParser<>("start_datafeed_request", a -> new StartDatafeedRequest((String)a[0]));
static {
PARSER.declareString(ConstructingObjectParser.constructorArg(), DatafeedConfig.ID);
PARSER.declareString(StartDatafeedRequest::setStart, START);
PARSER.declareString(StartDatafeedRequest::setEnd, END);
PARSER.declareString((params, val) ->
params.setTimeout(TimeValue.parseTimeValue(val, TIMEOUT.getPreferredName())), TIMEOUT);
}
private final String datafeedId;
private String start;
private String end;
private TimeValue timeout;
/**
* Create a new StartDatafeedRequest for the given DatafeedId
*
* @param datafeedId non-null existing Datafeed ID
*/
public StartDatafeedRequest(String datafeedId) {
this.datafeedId = Objects.requireNonNull(datafeedId, "[datafeed_id] must not be null");
}
public String getDatafeedId() {
return datafeedId;
}
public String getStart() {
return start;
}
/**
* The time that the datafeed should begin. This value is inclusive.
*
* If you specify a start value that is earlier than the timestamp of the latest processed record,
* the datafeed continues from 1 millisecond after the timestamp of the latest processed record.
*
* If you do not specify a start time and the datafeed is associated with a new job,
* the analysis starts from the earliest time for which data is available.
*
* @param start String representation of a timestamp; may be an epoch seconds, epoch millis or an ISO 8601 string
*/
public void setStart(String start) {
this.start = start;
}
public String getEnd() {
return end;
}
/**
* The time that the datafeed should end. This value is exclusive.
* If you do not specify an end time, the datafeed runs continuously.
*
* @param end String representation of a timestamp; may be an epoch seconds, epoch millis or an ISO 8601 string
*/
public void setEnd(String end) {
this.end = end;
}
public TimeValue getTimeout() {
return timeout;
}
/**
* Indicates how long to wait for the cluster to respond to the request.
*
* @param timeout TimeValue for how long to wait for a response from the cluster
*/
public void setTimeout(TimeValue timeout) {
this.timeout = timeout;
}
@Override
public ActionRequestValidationException validate() {
return null;
}
@Override
public int hashCode() {
return Objects.hash(datafeedId, start, end, timeout);
}
@Override
public boolean equals(Object obj) {
if (this == obj) {
return true;
}
if (obj == null || obj.getClass() != getClass()) {
return false;
}
StartDatafeedRequest other = (StartDatafeedRequest) obj;
return Objects.equals(datafeedId, other.datafeedId) &&
Objects.equals(start, other.start) &&
Objects.equals(end, other.end) &&
Objects.equals(timeout, other.timeout);
}
@Override
public XContentBuilder toXContent(XContentBuilder builder, Params params) throws IOException {
builder.startObject();
builder.field(DatafeedConfig.ID.getPreferredName(), datafeedId);
if (start != null) {
builder.field(START.getPreferredName(), start);
}
if (end != null) {
builder.field(END.getPreferredName(), end);
}
if (timeout != null) {
builder.field(TIMEOUT.getPreferredName(), timeout.getStringRep());
}
builder.endObject();
return builder;
}
}

View File

@ -0,0 +1,93 @@
/*
* Licensed to Elasticsearch under one or more contributor
* license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright
* ownership. Elasticsearch licenses this file to you under
* the Apache License, Version 2.0 (the "License"); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing,
* software distributed under the License is distributed on an
* "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
* KIND, either express or implied. See the License for the
* specific language governing permissions and limitations
* under the License.
*/
package org.elasticsearch.client.ml;
import org.elasticsearch.action.ActionResponse;
import org.elasticsearch.common.ParseField;
import org.elasticsearch.common.xcontent.ConstructingObjectParser;
import org.elasticsearch.common.xcontent.ToXContentObject;
import org.elasticsearch.common.xcontent.XContentBuilder;
import org.elasticsearch.common.xcontent.XContentParser;
import java.io.IOException;
import java.util.Objects;
/**
* Response indicating if the Machine Learning Datafeed is now started or not
*/
public class StartDatafeedResponse extends ActionResponse implements ToXContentObject {
private static final ParseField STARTED = new ParseField("started");
public static final ConstructingObjectParser<StartDatafeedResponse, Void> PARSER =
new ConstructingObjectParser<>(
"start_datafeed_response",
true,
(a) -> new StartDatafeedResponse((Boolean)a[0]));
static {
PARSER.declareBoolean(ConstructingObjectParser.constructorArg(), STARTED);
}
private final boolean started;
public StartDatafeedResponse(boolean started) {
this.started = started;
}
public static StartDatafeedResponse fromXContent(XContentParser parser) throws IOException {
return PARSER.parse(parser, null);
}
/**
* Has the Datafeed started or not
*
* @return boolean value indicating the Datafeed started status
*/
public boolean isStarted() {
return started;
}
@Override
public boolean equals(Object other) {
if (this == other) {
return true;
}
if (other == null || getClass() != other.getClass()) {
return false;
}
StartDatafeedResponse that = (StartDatafeedResponse) other;
return isStarted() == that.isStarted();
}
@Override
public int hashCode() {
return Objects.hash(isStarted());
}
@Override
public XContentBuilder toXContent(XContentBuilder builder, Params params) throws IOException {
builder.startObject();
builder.field(STARTED.getPreferredName(), started);
builder.endObject();
return builder;
}
}

View File

@ -44,6 +44,8 @@ import org.elasticsearch.client.ml.PostDataRequest;
import org.elasticsearch.client.ml.PutCalendarRequest; import org.elasticsearch.client.ml.PutCalendarRequest;
import org.elasticsearch.client.ml.PutDatafeedRequest; import org.elasticsearch.client.ml.PutDatafeedRequest;
import org.elasticsearch.client.ml.PutJobRequest; import org.elasticsearch.client.ml.PutJobRequest;
import org.elasticsearch.client.ml.StartDatafeedRequest;
import org.elasticsearch.client.ml.StartDatafeedRequestTests;
import org.elasticsearch.client.ml.UpdateJobRequest; import org.elasticsearch.client.ml.UpdateJobRequest;
import org.elasticsearch.client.ml.calendars.Calendar; import org.elasticsearch.client.ml.calendars.Calendar;
import org.elasticsearch.client.ml.calendars.CalendarTests; import org.elasticsearch.client.ml.calendars.CalendarTests;
@ -261,6 +263,19 @@ public class MLRequestConvertersTests extends ESTestCase {
assertEquals(Boolean.toString(true), request.getParameters().get("force")); assertEquals(Boolean.toString(true), request.getParameters().get("force"));
} }
public void testStartDatafeed() throws Exception {
String datafeedId = DatafeedConfigTests.randomValidDatafeedId();
StartDatafeedRequest datafeedRequest = StartDatafeedRequestTests.createRandomInstance(datafeedId);
Request request = MLRequestConverters.startDatafeed(datafeedRequest);
assertEquals(HttpPost.METHOD_NAME, request.getMethod());
assertEquals("/_xpack/ml/datafeeds/" + datafeedId + "/_start", request.getEndpoint());
try (XContentParser parser = createParser(JsonXContent.jsonXContent, request.getEntity().getContent())) {
StartDatafeedRequest parsedDatafeedRequest = StartDatafeedRequest.PARSER.apply(parser, null);
assertThat(parsedDatafeedRequest, equalTo(datafeedRequest));
}
}
public void testDeleteForecast() { public void testDeleteForecast() {
String jobId = randomAlphaOfLength(10); String jobId = randomAlphaOfLength(10);
DeleteForecastRequest deleteForecastRequest = new DeleteForecastRequest(jobId); DeleteForecastRequest deleteForecastRequest = new DeleteForecastRequest(jobId);

View File

@ -20,8 +20,12 @@ package org.elasticsearch.client;
import com.carrotsearch.randomizedtesting.generators.CodepointSetGenerator; import com.carrotsearch.randomizedtesting.generators.CodepointSetGenerator;
import org.elasticsearch.ElasticsearchStatusException; import org.elasticsearch.ElasticsearchStatusException;
import org.elasticsearch.action.admin.indices.create.CreateIndexRequest;
import org.elasticsearch.action.bulk.BulkRequest;
import org.elasticsearch.action.get.GetRequest; import org.elasticsearch.action.get.GetRequest;
import org.elasticsearch.action.get.GetResponse; import org.elasticsearch.action.get.GetResponse;
import org.elasticsearch.action.index.IndexRequest;
import org.elasticsearch.action.support.WriteRequest;
import org.elasticsearch.action.support.master.AcknowledgedResponse; import org.elasticsearch.action.support.master.AcknowledgedResponse;
import org.elasticsearch.client.ml.CloseJobRequest; import org.elasticsearch.client.ml.CloseJobRequest;
import org.elasticsearch.client.ml.CloseJobResponse; import org.elasticsearch.client.ml.CloseJobResponse;
@ -51,6 +55,8 @@ import org.elasticsearch.client.ml.PutDatafeedRequest;
import org.elasticsearch.client.ml.PutDatafeedResponse; import org.elasticsearch.client.ml.PutDatafeedResponse;
import org.elasticsearch.client.ml.PutJobRequest; import org.elasticsearch.client.ml.PutJobRequest;
import org.elasticsearch.client.ml.PutJobResponse; import org.elasticsearch.client.ml.PutJobResponse;
import org.elasticsearch.client.ml.StartDatafeedRequest;
import org.elasticsearch.client.ml.StartDatafeedResponse;
import org.elasticsearch.client.ml.UpdateJobRequest; import org.elasticsearch.client.ml.UpdateJobRequest;
import org.elasticsearch.client.ml.calendars.Calendar; import org.elasticsearch.client.ml.calendars.Calendar;
import org.elasticsearch.client.ml.calendars.CalendarTests; import org.elasticsearch.client.ml.calendars.CalendarTests;
@ -63,6 +69,7 @@ import org.elasticsearch.client.ml.job.config.JobState;
import org.elasticsearch.client.ml.job.config.JobUpdate; import org.elasticsearch.client.ml.job.config.JobUpdate;
import org.elasticsearch.client.ml.job.stats.JobStats; import org.elasticsearch.client.ml.job.stats.JobStats;
import org.elasticsearch.common.unit.TimeValue; import org.elasticsearch.common.unit.TimeValue;
import org.elasticsearch.common.xcontent.XContentType;
import org.elasticsearch.rest.RestStatus; import org.elasticsearch.rest.RestStatus;
import org.junit.After; import org.junit.After;
@ -416,6 +423,80 @@ public class MachineLearningIT extends ESRestHighLevelClientTestCase {
assertTrue(response.isAcknowledged()); assertTrue(response.isAcknowledged());
} }
public void testStartDatafeed() throws Exception {
String jobId = "test-start-datafeed";
String indexName = "start_data_1";
// Set up the index and docs
CreateIndexRequest createIndexRequest = new CreateIndexRequest(indexName);
createIndexRequest.mapping("doc", "timestamp", "type=date", "total", "type=long");
highLevelClient().indices().create(createIndexRequest, RequestOptions.DEFAULT);
BulkRequest bulk = new BulkRequest();
bulk.setRefreshPolicy(WriteRequest.RefreshPolicy.IMMEDIATE);
long now = System.currentTimeMillis();
long oneDayAgo = now - 86400000;
int i = 0;
long dayAgoCopy = oneDayAgo;
while(dayAgoCopy < now) {
IndexRequest doc = new IndexRequest();
doc.index(indexName);
doc.type("doc");
doc.id("id" + i);
doc.source("{\"total\":" +randomInt(1000) + ",\"timestamp\":"+ dayAgoCopy +"}", XContentType.JSON);
bulk.add(doc);
dayAgoCopy += 1000000;
i++;
}
highLevelClient().bulk(bulk, RequestOptions.DEFAULT);
final long totalDocCount = i;
// create the job and the datafeed
Job job = buildJob(jobId);
MachineLearningClient machineLearningClient = highLevelClient().machineLearning();
machineLearningClient.putJob(new PutJobRequest(job), RequestOptions.DEFAULT);
machineLearningClient.openJob(new OpenJobRequest(jobId), RequestOptions.DEFAULT);
String datafeedId = jobId + "-feed";
DatafeedConfig datafeed = DatafeedConfig.builder(datafeedId, jobId)
.setIndices(indexName)
.setQueryDelay(TimeValue.timeValueSeconds(1))
.setTypes(Arrays.asList("doc"))
.setFrequency(TimeValue.timeValueSeconds(1)).build();
machineLearningClient.putDatafeed(new PutDatafeedRequest(datafeed), RequestOptions.DEFAULT);
StartDatafeedRequest startDatafeedRequest = new StartDatafeedRequest(datafeedId);
startDatafeedRequest.setStart(String.valueOf(oneDayAgo));
// Should only process two documents
startDatafeedRequest.setEnd(String.valueOf(oneDayAgo + 2000000));
StartDatafeedResponse response = execute(startDatafeedRequest,
machineLearningClient::startDatafeed,
machineLearningClient::startDatafeedAsync);
assertTrue(response.isStarted());
assertBusy(() -> {
JobStats stats = machineLearningClient.getJobStats(new GetJobStatsRequest(jobId), RequestOptions.DEFAULT).jobStats().get(0);
assertEquals(2L, stats.getDataCounts().getInputRecordCount());
assertEquals(JobState.CLOSED, stats.getState());
}, 30, TimeUnit.SECONDS);
machineLearningClient.openJob(new OpenJobRequest(jobId), RequestOptions.DEFAULT);
StartDatafeedRequest wholeDataFeed = new StartDatafeedRequest(datafeedId);
// Process all documents and end the stream
wholeDataFeed.setEnd(String.valueOf(now));
StartDatafeedResponse wholeResponse = execute(wholeDataFeed,
machineLearningClient::startDatafeed,
machineLearningClient::startDatafeedAsync);
assertTrue(wholeResponse.isStarted());
assertBusy(() -> {
JobStats stats = machineLearningClient.getJobStats(new GetJobStatsRequest(jobId), RequestOptions.DEFAULT).jobStats().get(0);
assertEquals(totalDocCount, stats.getDataCounts().getInputRecordCount());
assertEquals(JobState.CLOSED, stats.getState());
}, 30, TimeUnit.SECONDS);
}
public void testDeleteForecast() throws Exception { public void testDeleteForecast() throws Exception {
String jobId = "test-delete-forecast"; String jobId = "test-delete-forecast";

View File

@ -20,6 +20,7 @@ package org.elasticsearch.client.documentation;
import org.elasticsearch.action.ActionListener; import org.elasticsearch.action.ActionListener;
import org.elasticsearch.action.LatchedActionListener; import org.elasticsearch.action.LatchedActionListener;
import org.elasticsearch.action.admin.indices.create.CreateIndexRequest;
import org.elasticsearch.action.bulk.BulkRequest; import org.elasticsearch.action.bulk.BulkRequest;
import org.elasticsearch.action.get.GetRequest; import org.elasticsearch.action.get.GetRequest;
import org.elasticsearch.action.get.GetResponse; import org.elasticsearch.action.get.GetResponse;
@ -70,6 +71,8 @@ import org.elasticsearch.client.ml.PutDatafeedRequest;
import org.elasticsearch.client.ml.PutDatafeedResponse; import org.elasticsearch.client.ml.PutDatafeedResponse;
import org.elasticsearch.client.ml.PutJobRequest; import org.elasticsearch.client.ml.PutJobRequest;
import org.elasticsearch.client.ml.PutJobResponse; import org.elasticsearch.client.ml.PutJobResponse;
import org.elasticsearch.client.ml.StartDatafeedRequest;
import org.elasticsearch.client.ml.StartDatafeedResponse;
import org.elasticsearch.client.ml.UpdateJobRequest; import org.elasticsearch.client.ml.UpdateJobRequest;
import org.elasticsearch.client.ml.calendars.Calendar; import org.elasticsearch.client.ml.calendars.Calendar;
import org.elasticsearch.client.ml.datafeed.ChunkingConfig; import org.elasticsearch.client.ml.datafeed.ChunkingConfig;
@ -703,6 +706,69 @@ public class MlClientDocumentationIT extends ESRestHighLevelClientTestCase {
} }
} }
public void testStartDatafeed() throws Exception {
RestHighLevelClient client = highLevelClient();
Job job = MachineLearningIT.buildJob("start-datafeed-job");
client.machineLearning().putJob(new PutJobRequest(job), RequestOptions.DEFAULT);
String datafeedId = job.getId() + "-feed";
String indexName = "start_data_2";
CreateIndexRequest createIndexRequest = new CreateIndexRequest(indexName);
createIndexRequest.mapping("doc", "timestamp", "type=date", "total", "type=long");
highLevelClient().indices().create(createIndexRequest, RequestOptions.DEFAULT);
DatafeedConfig datafeed = DatafeedConfig.builder(datafeedId, job.getId())
.setTypes(Arrays.asList("doc"))
.setIndices(indexName)
.build();
client.machineLearning().putDatafeed(new PutDatafeedRequest(datafeed), RequestOptions.DEFAULT);
client.machineLearning().openJob(new OpenJobRequest(job.getId()), RequestOptions.DEFAULT);
{
//tag::x-pack-ml-start-datafeed-request
StartDatafeedRequest request = new StartDatafeedRequest(datafeedId); // <1>
//end::x-pack-ml-start-datafeed-request
//tag::x-pack-ml-start-datafeed-request-options
request.setEnd("2018-08-21T00:00:00Z"); // <1>
request.setStart("2018-08-20T00:00:00Z"); // <2>
request.setTimeout(TimeValue.timeValueMinutes(10)); // <3>
//end::x-pack-ml-start-datafeed-request-options
//tag::x-pack-ml-start-datafeed-execute
StartDatafeedResponse response = client.machineLearning().startDatafeed(request, RequestOptions.DEFAULT);
boolean started = response.isStarted(); // <1>
//end::x-pack-ml-start-datafeed-execute
assertTrue(started);
}
{
StartDatafeedRequest request = new StartDatafeedRequest(datafeedId);
// tag::x-pack-ml-start-datafeed-listener
ActionListener<StartDatafeedResponse> listener = new ActionListener<StartDatafeedResponse>() {
@Override
public void onResponse(StartDatafeedResponse response) {
// <1>
}
@Override
public void onFailure(Exception e) {
// <2>
}
};
// end::x-pack-ml-start-datafeed-listener
// Replace the empty listener by a blocking listener in test
final CountDownLatch latch = new CountDownLatch(1);
listener = new LatchedActionListener<>(listener, latch);
// tag::x-pack-ml-start-datafeed-execute-async
client.machineLearning().startDatafeedAsync(request, RequestOptions.DEFAULT, listener); // <1>
// end::x-pack-ml-start-datafeed-execute-async
assertTrue(latch.await(30L, TimeUnit.SECONDS));
}
}
public void testGetBuckets() throws IOException, InterruptedException { public void testGetBuckets() throws IOException, InterruptedException {
RestHighLevelClient client = highLevelClient(); RestHighLevelClient client = highLevelClient();

View File

@ -0,0 +1,61 @@
/*
* Licensed to Elasticsearch under one or more contributor
* license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright
* ownership. Elasticsearch licenses this file to you under
* the Apache License, Version 2.0 (the "License"); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing,
* software distributed under the License is distributed on an
* "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
* KIND, either express or implied. See the License for the
* specific language governing permissions and limitations
* under the License.
*/
package org.elasticsearch.client.ml;
import org.elasticsearch.client.ml.datafeed.DatafeedConfigTests;
import org.elasticsearch.common.unit.TimeValue;
import org.elasticsearch.common.xcontent.XContentParser;
import org.elasticsearch.test.AbstractXContentTestCase;
import java.io.IOException;
public class StartDatafeedRequestTests extends AbstractXContentTestCase<StartDatafeedRequest> {
public static StartDatafeedRequest createRandomInstance(String datafeedId) {
StartDatafeedRequest request = new StartDatafeedRequest(datafeedId);
if (randomBoolean()) {
request.setStart(String.valueOf(randomLongBetween(1, 1000)));
}
if (randomBoolean()) {
request.setEnd(String.valueOf(randomLongBetween(1, 1000)));
}
if (randomBoolean()) {
request.setTimeout(TimeValue.timeValueMinutes(randomLongBetween(1, 1000)));
}
return request;
}
@Override
protected StartDatafeedRequest createTestInstance() {
String datafeedId = DatafeedConfigTests.randomValidDatafeedId();
return createRandomInstance(datafeedId);
}
@Override
protected StartDatafeedRequest doParseInstance(XContentParser parser) throws IOException {
return StartDatafeedRequest.PARSER.parse(parser, null);
}
@Override
protected boolean supportsUnknownFields() {
return false;
}
}

View File

@ -0,0 +1,42 @@
/*
* Licensed to Elasticsearch under one or more contributor
* license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright
* ownership. Elasticsearch licenses this file to you under
* the Apache License, Version 2.0 (the "License"); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing,
* software distributed under the License is distributed on an
* "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
* KIND, either express or implied. See the License for the
* specific language governing permissions and limitations
* under the License.
*/
package org.elasticsearch.client.ml;
import org.elasticsearch.common.xcontent.XContentParser;
import org.elasticsearch.test.AbstractXContentTestCase;
import java.io.IOException;
public class StartDatafeedResponseTests extends AbstractXContentTestCase<StartDatafeedResponse> {
@Override
protected StartDatafeedResponse createTestInstance() {
return new StartDatafeedResponse(randomBoolean());
}
@Override
protected StartDatafeedResponse doParseInstance(XContentParser parser) throws IOException {
return StartDatafeedResponse.fromXContent(parser);
}
@Override
protected boolean supportsUnknownFields() {
return true;
}
}

View File

@ -0,0 +1,71 @@
[[java-rest-high-x-pack-ml-start-datafeed]]
=== Start Datafeed API
The Start Datafeed API provides the ability to start a {ml} datafeed in the cluster.
It accepts a `StartDatafeedRequest` object and responds
with a `StartDatafeedResponse` object.
[[java-rest-high-x-pack-ml-start-datafeed-request]]
==== Start Datafeed Request
A `StartDatafeedRequest` object is created referencing a non-null `datafeedId`.
All other fields are optional for the request.
["source","java",subs="attributes,callouts,macros"]
--------------------------------------------------
include-tagged::{doc-tests}/MlClientDocumentationIT.java[x-pack-ml-start-datafeed-request]
--------------------------------------------------
<1> Constructing a new request referencing an existing `datafeedId`
==== Optional Arguments
The following arguments are optional.
["source","java",subs="attributes,callouts,macros"]
--------------------------------------------------
include-tagged::{doc-tests}/MlClientDocumentationIT.java[x-pack-ml-start-datafeed-request-options]
--------------------------------------------------
<1> Set when the datafeed should end, the value is exclusive.
May be an epoch seconds, epoch millis or an ISO 8601 string.
"now" is a special value that indicates the current time.
If you do not specify an end time, the datafeed runs continuously.
<2> Set when the datafeed should start, the value is inclusive.
May be an epoch seconds, epoch millis or an ISO 8601 string.
If you do not specify a start time and the datafeed is associated with a new job,
the analysis starts from the earliest time for which data is available.
<3> Set the timeout for the request
[[java-rest-high-x-pack-ml-start-datafeed-execution]]
==== Execution
The request can be executed through the `MachineLearningClient` contained
in the `RestHighLevelClient` object, accessed via the `machineLearningClient()` method.
["source","java",subs="attributes,callouts,macros"]
--------------------------------------------------
include-tagged::{doc-tests}/MlClientDocumentationIT.java[x-pack-ml-start-datafeed-execute]
--------------------------------------------------
<1> Did the datafeed successfully start?
[[java-rest-high-x-pack-ml-start-datafeed-execution-async]]
==== Asynchronous Execution
The request can also be executed asynchronously:
["source","java",subs="attributes,callouts,macros"]
--------------------------------------------------
include-tagged::{doc-tests}/MlClientDocumentationIT.java[x-pack-ml-start-datafeed-execute-async]
--------------------------------------------------
<1> The `StartDatafeedRequest` to execute and the `ActionListener` to use when
the execution completes
The method does not block and returns immediately. The passed `ActionListener` is used
to notify the caller of completion. A typical `ActionListener` for `StartDatafeedResponse` may
look like
["source","java",subs="attributes,callouts,macros"]
--------------------------------------------------
include-tagged::{doc-tests}/MlClientDocumentationIT.java[x-pack-ml-start-datafeed-listener]
--------------------------------------------------
<1> `onResponse` is called back when the action is completed successfully
<2> `onFailure` is called back when some unexpected error occurs

View File

@ -225,6 +225,7 @@ The Java High Level REST Client supports the following Machine Learning APIs:
* <<java-rest-high-x-pack-ml-put-datafeed>> * <<java-rest-high-x-pack-ml-put-datafeed>>
* <<java-rest-high-x-pack-ml-get-datafeed>> * <<java-rest-high-x-pack-ml-get-datafeed>>
* <<java-rest-high-x-pack-ml-delete-datafeed>> * <<java-rest-high-x-pack-ml-delete-datafeed>>
* <<java-rest-high-x-pack-ml-start-datafeed>>
* <<java-rest-high-x-pack-ml-forecast-job>> * <<java-rest-high-x-pack-ml-forecast-job>>
* <<java-rest-high-x-pack-ml-delete-forecast>> * <<java-rest-high-x-pack-ml-delete-forecast>>
* <<java-rest-high-x-pack-ml-get-buckets>> * <<java-rest-high-x-pack-ml-get-buckets>>
@ -247,6 +248,7 @@ include::ml/flush-job.asciidoc[]
include::ml/put-datafeed.asciidoc[] include::ml/put-datafeed.asciidoc[]
include::ml/get-datafeed.asciidoc[] include::ml/get-datafeed.asciidoc[]
include::ml/delete-datafeed.asciidoc[] include::ml/delete-datafeed.asciidoc[]
include::ml/start-datafeed.asciidoc[]
include::ml/get-job-stats.asciidoc[] include::ml/get-job-stats.asciidoc[]
include::ml/forecast-job.asciidoc[] include::ml/forecast-job.asciidoc[]
include::ml/delete-forecast.asciidoc[] include::ml/delete-forecast.asciidoc[]