Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
@@ -0,0 +1,244 @@
/*
* Copyright Elasticsearch B.V. and/or licensed to Elasticsearch B.V. under one
* or more contributor license agreements. Licensed under the Elastic License;
* you may not use this file except in compliance with the Elastic License.
*/
package org.elasticsearch.xpack.core.ml.annotations;

import org.elasticsearch.common.ParseField;
import org.elasticsearch.common.io.stream.StreamInput;
import org.elasticsearch.common.io.stream.StreamOutput;
import org.elasticsearch.common.io.stream.Writeable;
import org.elasticsearch.common.xcontent.ObjectParser;
import org.elasticsearch.common.xcontent.ToXContentObject;
import org.elasticsearch.common.xcontent.XContentBuilder;
import org.elasticsearch.xpack.core.ml.job.config.Job;
import org.elasticsearch.xpack.core.ml.utils.time.TimeUtils;

import java.io.IOException;
import java.util.Date;
import java.util.Objects;

public class Annotation implements ToXContentObject, Writeable {

public static final ParseField ANNOTATION = new ParseField("annotation");
public static final ParseField CREATE_TIME = new ParseField("create_time");
public static final ParseField CREATE_USERNAME = new ParseField("create_username");
public static final ParseField TIMESTAMP = new ParseField("timestamp");
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

is timestamp the partner of end_timestamp? Should it be start_timestamp

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The reason timestamp was chosen is that originally we'd intended annotations to be stored in the results index. Then we decided to switch to a separate index, but didn't change the field names.

I think it's too late to change now as it's feature freeze day and
https://github.com/elastic/kibana/pull/26034/files#diff-c5c6ac3dbb0e7c91b6d127aa06121b2cR7 is already merged.

But one thing I just noticed is that end_timestamp is optional in the UI code, so actually that makes timestamp seem not as bad, as it could either be the start or the sole time. But I need to update this PR to account for end_timestamp being optional.

public static final ParseField END_TIMESTAMP = new ParseField("end_timestamp");
public static final ParseField MODIFIED_TIME = new ParseField("modified_time");
public static final ParseField MODIFIED_USERNAME = new ParseField("modified_username");
public static final ParseField TYPE = new ParseField("type");

public static final ObjectParser<Annotation, Void> PARSER = new ObjectParser<>(TYPE.getPreferredName(), true, Annotation::new);

static {
PARSER.declareString(Annotation::setAnnotation, ANNOTATION);
PARSER.declareField(Annotation::setCreateTime,
p -> TimeUtils.parseTimeField(p, CREATE_TIME.getPreferredName()), CREATE_TIME, ObjectParser.ValueType.VALUE);
PARSER.declareString(Annotation::setCreateUsername, CREATE_USERNAME);
PARSER.declareField(Annotation::setTimestamp,
p -> TimeUtils.parseTimeField(p, TIMESTAMP.getPreferredName()), TIMESTAMP, ObjectParser.ValueType.VALUE);
PARSER.declareField(Annotation::setEndTimestamp,
p -> TimeUtils.parseTimeField(p, END_TIMESTAMP.getPreferredName()), END_TIMESTAMP, ObjectParser.ValueType.VALUE);
PARSER.declareString(Annotation::setJobId, Job.ID);
PARSER.declareField(Annotation::setModifiedTime,
p -> TimeUtils.parseTimeField(p, MODIFIED_TIME.getPreferredName()), MODIFIED_TIME, ObjectParser.ValueType.VALUE);
PARSER.declareString(Annotation::setModifiedUsername, MODIFIED_USERNAME);
PARSER.declareString(Annotation::setType, TYPE);
}

private String annotation;
private Date createTime;
private String createUsername;
private Date timestamp;
private Date endTimestamp;
/**
* Unlike most ML classes, this may be <code>null</code> or wildcarded
*/
private String jobId;
private Date modifiedTime;
private String modifiedUsername;
private String type;

private Annotation() {
}

public Annotation(String annotation, Date createTime, String createUsername, Date timestamp, Date endTimestamp, String jobId,
Date modifiedTime, String modifiedUsername, String type) {
this.annotation = Objects.requireNonNull(annotation);
this.createTime = Objects.requireNonNull(createTime);
this.createUsername = Objects.requireNonNull(createUsername);
this.timestamp = Objects.requireNonNull(timestamp);
this.endTimestamp = endTimestamp;
this.jobId = jobId;
this.modifiedTime = modifiedTime;
this.modifiedUsername = modifiedUsername;
this.type = Objects.requireNonNull(type);
}

public Annotation(StreamInput in) throws IOException {
annotation = in.readString();
createTime = new Date(in.readLong());
createUsername = in.readString();
timestamp = new Date(in.readLong());
if (in.readBoolean()) {
endTimestamp = new Date(in.readLong());
} else {
endTimestamp = null;
}
jobId = in.readOptionalString();
if (in.readBoolean()) {
modifiedTime = new Date(in.readLong());
} else {
modifiedTime = null;
}
modifiedUsername = in.readOptionalString();
type = in.readString();
}

@Override
public void writeTo(StreamOutput out) throws IOException {
out.writeString(annotation);
out.writeLong(createTime.getTime());
out.writeString(createUsername);
out.writeLong(timestamp.getTime());
if (endTimestamp != null) {
out.writeBoolean(true);
out.writeLong(endTimestamp.getTime());
} else {
out.writeBoolean(false);
}
out.writeOptionalString(jobId);
if (modifiedTime != null) {
out.writeBoolean(true);
out.writeLong(modifiedTime.getTime());
} else {
out.writeBoolean(false);
}
out.writeOptionalString(modifiedUsername);
out.writeString(type);

}

public String getAnnotation() {
return annotation;
}

public void setAnnotation(String annotation) {
this.annotation = Objects.requireNonNull(annotation);
}

public Date getCreateTime() {
return createTime;
}

public void setCreateTime(Date createTime) {
this.createTime = Objects.requireNonNull(createTime);
}

public String getCreateUsername() {
return createUsername;
}

public void setCreateUsername(String createUsername) {
this.createUsername = Objects.requireNonNull(createUsername);
}

public Date getTimestamp() {
return timestamp;
}

public void setTimestamp(Date timestamp) {
this.timestamp = Objects.requireNonNull(timestamp);
}

public Date getEndTimestamp() {
return endTimestamp;
}

public void setEndTimestamp(Date endTimestamp) {
this.endTimestamp = endTimestamp;
}

public String getJobId() {
return jobId;
}

public void setJobId(String jobId) {
this.jobId = jobId;
}

public Date getModifiedTime() {
return modifiedTime;
}

public void setModifiedTime(Date modifiedTime) {
this.modifiedTime = modifiedTime;
}

public String getModifiedUsername() {
return modifiedUsername;
}

public void setModifiedUsername(String modifiedUsername) {
this.modifiedUsername = modifiedUsername;
}

public String getType() {
return type;
}

public void setType(String type) {
this.type = Objects.requireNonNull(type);
}

@Override
public XContentBuilder toXContent(XContentBuilder builder, Params params) throws IOException {
builder.startObject();
builder.field(ANNOTATION.getPreferredName(), annotation);
builder.timeField(CREATE_TIME.getPreferredName(), CREATE_TIME.getPreferredName() + "_string", createTime.getTime());
builder.field(CREATE_USERNAME.getPreferredName(), createUsername);
builder.timeField(TIMESTAMP.getPreferredName(), TIMESTAMP.getPreferredName() + "_string", timestamp.getTime());
if (endTimestamp != null) {
builder.timeField(END_TIMESTAMP.getPreferredName(), END_TIMESTAMP.getPreferredName() + "_string", endTimestamp.getTime());
}
if (jobId != null) {
builder.field(Job.ID.getPreferredName(), jobId);
}
if (modifiedTime != null) {
builder.timeField(MODIFIED_TIME.getPreferredName(), MODIFIED_TIME.getPreferredName() + "_string", modifiedTime.getTime());
}
if (modifiedUsername != null) {
builder.field(MODIFIED_USERNAME.getPreferredName(), modifiedUsername);
}
builder.field(TYPE.getPreferredName(), type);
builder.endObject();
return builder;
}

@Override
public int hashCode() {
return Objects.hash(annotation, createTime, createUsername, timestamp, endTimestamp, jobId, modifiedTime, modifiedUsername, type);
}

@Override
public boolean equals(Object obj) {
if (obj == this) {
return true;
}
if (obj == null || getClass() != obj.getClass()) {
return false;
}
Annotation other = (Annotation) obj;
return Objects.equals(annotation, other.annotation) &&
Objects.equals(createTime, other.createTime) &&
Objects.equals(createUsername, other.createUsername) &&
Objects.equals(timestamp, other.timestamp) &&
Objects.equals(endTimestamp, other.endTimestamp) &&
Objects.equals(jobId, other.jobId) &&
Objects.equals(modifiedTime, other.modifiedTime) &&
Objects.equals(modifiedUsername, other.modifiedUsername) &&
Objects.equals(type, other.type);
}
}
Original file line number Diff line number Diff line change
@@ -0,0 +1,147 @@
/*
* Copyright Elasticsearch B.V. and/or licensed to Elasticsearch B.V. under one
* or more contributor license agreements. Licensed under the Elastic License;
* you may not use this file except in compliance with the Elastic License.
*/
package org.elasticsearch.xpack.core.ml.annotations;

import org.elasticsearch.ResourceAlreadyExistsException;
import org.elasticsearch.action.ActionListener;
import org.elasticsearch.action.admin.indices.alias.IndicesAliasesRequest;
import org.elasticsearch.action.admin.indices.create.CreateIndexRequest;
import org.elasticsearch.action.admin.indices.create.CreateIndexResponse;
import org.elasticsearch.action.support.master.AcknowledgedResponse;
import org.elasticsearch.client.Client;
import org.elasticsearch.cluster.ClusterState;
import org.elasticsearch.cluster.metadata.AliasOrIndex;
import org.elasticsearch.cluster.metadata.IndexMetaData;
import org.elasticsearch.cluster.routing.UnassignedInfo;
import org.elasticsearch.common.settings.Settings;
import org.elasticsearch.common.unit.TimeValue;
import org.elasticsearch.common.xcontent.XContentBuilder;
import org.elasticsearch.xpack.core.ml.MachineLearningField;
import org.elasticsearch.xpack.core.ml.job.config.Job;
import org.elasticsearch.xpack.core.ml.job.persistence.ElasticsearchMappings;

import java.io.IOException;
import java.util.SortedMap;

import static org.elasticsearch.common.xcontent.XContentFactory.jsonBuilder;
import static org.elasticsearch.xpack.core.ClientHelper.ML_ORIGIN;
import static org.elasticsearch.xpack.core.ClientHelper.executeAsyncWithOrigin;

public class AnnotationIndex {

public static final String READ_ALIAS_NAME = ".ml-annotations-read";
public static final String WRITE_ALIAS_NAME = ".ml-annotations-write";
// Exposed for testing, but always use the aliases in non-test code
public static final String INDEX_NAME = ".ml-annotations-6";

/**
* Create the .ml-annotations index with correct mappings.
* This index is read and written by the UI results views,
* so needs to exist when there might be ML results to view.
*/
public static void createAnnotationsIndex(Settings settings, Client client, ClusterState state,
final ActionListener<Boolean> finalListener) {

final ActionListener<Boolean> createAliasListener = ActionListener.wrap(success -> {
final IndicesAliasesRequest request = client.admin().indices().prepareAliases()
.addAlias(INDEX_NAME, READ_ALIAS_NAME)
.addAlias(INDEX_NAME, WRITE_ALIAS_NAME).request();
executeAsyncWithOrigin(client.threadPool().getThreadContext(), ML_ORIGIN, request,
ActionListener.<AcknowledgedResponse>wrap(r -> finalListener.onResponse(r.isAcknowledged()), finalListener::onFailure),
client.admin().indices()::aliases);
}, finalListener::onFailure);

// Only create the index or aliases if some other ML index exists - saves clutter if ML is never used.

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: the method is called createAnnotationsIndex, it would be nice to reflect this, e.g. maybeCreate... or move the "maybe"-part out of createAnnotationsIndex

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This PR is already merged, but I'll add maybe if there's a followup PR for any reason

SortedMap<String, AliasOrIndex> mlLookup = state.getMetaData().getAliasAndIndexLookup().tailMap(".ml");
if (mlLookup.isEmpty() == false && mlLookup.firstKey().startsWith(".ml")) {

// Create the annotations index if it doesn't exist already.
if (mlLookup.containsKey(INDEX_NAME) == false) {

final TimeValue delayedNodeTimeOutSetting;
// Whether we are using native process is a good way to detect whether we are in dev / test mode:
if (MachineLearningField.AUTODETECT_PROCESS.get(settings)) {
delayedNodeTimeOutSetting = UnassignedInfo.INDEX_DELAYED_NODE_LEFT_TIMEOUT_SETTING.get(settings);
} else {
delayedNodeTimeOutSetting = TimeValue.ZERO;
}

CreateIndexRequest createIndexRequest = new CreateIndexRequest(INDEX_NAME);
try (XContentBuilder annotationsMapping = AnnotationIndex.annotationsMapping()) {
createIndexRequest.mapping(ElasticsearchMappings.DOC_TYPE, annotationsMapping);
createIndexRequest.settings(Settings.builder()
.put(IndexMetaData.SETTING_AUTO_EXPAND_REPLICAS, "0-1")
.put(IndexMetaData.SETTING_NUMBER_OF_SHARDS, "1")
.put(UnassignedInfo.INDEX_DELAYED_NODE_LEFT_TIMEOUT_SETTING.getKey(), delayedNodeTimeOutSetting));

executeAsyncWithOrigin(client.threadPool().getThreadContext(), ML_ORIGIN, createIndexRequest,
ActionListener.<CreateIndexResponse>wrap(
r -> createAliasListener.onResponse(r.isAcknowledged()),
e -> {
// Possible that the index was created while the request was executing,
// so we need to handle that possibility
if (e instanceof ResourceAlreadyExistsException) {
// Create the alias
createAliasListener.onResponse(true);
} else {
finalListener.onFailure(e);
}
}
), client.admin().indices()::create);
} catch (IOException e) {
finalListener.onFailure(e);
}
return;
}

// Recreate the aliases if they've gone even though the index still exists.
if (mlLookup.containsKey(READ_ALIAS_NAME) == false || mlLookup.containsKey(WRITE_ALIAS_NAME) == false) {
createAliasListener.onResponse(true);
return;
}
}

// Nothing to do, but respond to the listener
finalListener.onResponse(false);
}

public static XContentBuilder annotationsMapping() throws IOException {
return jsonBuilder()
.startObject()
.startObject(ElasticsearchMappings.DOC_TYPE)
.startObject(ElasticsearchMappings.PROPERTIES)
.startObject(Annotation.ANNOTATION.getPreferredName())
.field(ElasticsearchMappings.TYPE, ElasticsearchMappings.TEXT)
.endObject()
.startObject(Annotation.CREATE_TIME.getPreferredName())
.field(ElasticsearchMappings.TYPE, ElasticsearchMappings.DATE)
.endObject()
.startObject(Annotation.CREATE_USERNAME.getPreferredName())
.field(ElasticsearchMappings.TYPE, ElasticsearchMappings.KEYWORD)
.endObject()
.startObject(Annotation.TIMESTAMP.getPreferredName())
.field(ElasticsearchMappings.TYPE, ElasticsearchMappings.DATE)
.endObject()
.startObject(Annotation.END_TIMESTAMP.getPreferredName())
.field(ElasticsearchMappings.TYPE, ElasticsearchMappings.DATE)
.endObject()
.startObject(Job.ID.getPreferredName())
.field(ElasticsearchMappings.TYPE, ElasticsearchMappings.KEYWORD)
.endObject()
.startObject(Annotation.MODIFIED_TIME.getPreferredName())
.field(ElasticsearchMappings.TYPE, ElasticsearchMappings.DATE)
.endObject()
.startObject(Annotation.MODIFIED_USERNAME.getPreferredName())
.field(ElasticsearchMappings.TYPE, ElasticsearchMappings.KEYWORD)
.endObject()
.startObject(Annotation.TYPE.getPreferredName())
.field(ElasticsearchMappings.TYPE, ElasticsearchMappings.KEYWORD)
.endObject()
.endObject()
.endObject()
.endObject();
}
}
Loading