[ML] Wait for controller to respond to commands

This change makes threads that send a command to the ML
controller process wait for it to respond to the command.
Previously such threads would block until the command was
sent, but not until it was actioned.  This was on the
assumption that the sort of commands being sent would be
actioned almost instantaneously, but that assumption has
been shown to be false when anti-malware software is
running.

Relates elastic/ml-cpp#1520
Fixes elastic#62823

Author: droberts195

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/MachineLearning.java

@@ -625,7 +625,8 @@ public Collection<Object> createComponents(Client client, ClusterService cluster
         AnalyticsProcessFactory<MemoryUsageEstimationResult> memoryEstimationProcessFactory;
         if (MachineLearningField.AUTODETECT_PROCESS.get(settings)) {
             try {
-                NativeController nativeController = NativeController.makeNativeController(clusterService.getNodeName(), environment);
+                NativeController nativeController =
+                    NativeController.makeNativeController(clusterService.getNodeName(), environment, xContentRegistry);
                 autodetectProcessFactory = new NativeAutodetectProcessFactory(
                     environment,
                     settings,

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/dataframe/process/AnalyticsBuilder.java

@@ -52,7 +52,7 @@ public AnalyticsBuilder performMemoryUsageEstimationOnly() {
         return this;
     }
 
-    public void build() throws IOException {
+    public void build() throws IOException, InterruptedException {
         List<String> command = buildAnalyticsCommand();
         processPipes.addArgs(command);
         nativeController.startProcess(command);

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/dataframe/process/AnalyticsProcessManager.java

@@ -83,7 +83,7 @@ public AnalyticsProcessManager(Settings settings,
         this(
             settings,
             client,
-            threadPool.generic(),
+            threadPool.executor(MachineLearning.UTILITY_THREAD_POOL_NAME),
             threadPool.executor(MachineLearning.JOB_COMMS_THREAD_POOL_NAME),
             analyticsProcessFactory,
             auditor,
@@ -451,7 +451,7 @@ synchronized void stop() {
             }
             if (process.get() != null) {
                 try {
-                    process.get().kill();
+                    process.get().kill(true);
                 } catch (IOException e) {
                     LOGGER.error(new ParameterizedMessage("[{}] Failed to kill process", config.getId()), e);
                 }

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/dataframe/process/NativeAnalyticsProcessFactory.java

@@ -116,6 +116,9 @@ private void createNativeProcess(String jobId, AnalyticsProcessConfig analyticsP
             new AnalyticsBuilder(env::tmpFile, nativeController, processPipes, analyticsProcessConfig, filesToDelete);
         try {
             analyticsBuilder.build();
+        } catch (InterruptedException e) {
+            Thread.currentThread().interrupt();
+            LOGGER.warn("Interrupted while launching data frame analytics process for job " + jobId);
         } catch (IOException e) {
             String msg = "Failed to launch data frame analytics process for job " + jobId;
             LOGGER.error(msg);

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/dataframe/process/NativeMemoryUsageEstimationProcessFactory.java

@@ -102,6 +102,9 @@ private void createNativeProcess(String jobId, AnalyticsProcessConfig analyticsP
                 .performMemoryUsageEstimationOnly();
         try {
             analyticsBuilder.build();
+        } catch (InterruptedException e) {
+            Thread.currentThread().interrupt();
+            LOGGER.warn("Interrupted while launching data frame analytics memory usage estimation process for job " + jobId);
         } catch (IOException e) {
             String msg = "Failed to launch data frame analytics memory usage estimation process for job " + jobId;
             LOGGER.error(msg);

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/job/process/autodetect/AutodetectBuilder.java

@@ -169,7 +169,7 @@ public AutodetectBuilder scheduledEvents(List<ScheduledEvent> scheduledEvents) {
     /**
      * Requests that the controller daemon start an autodetect process.
      */
-    public void build() throws IOException {
+    public void build() throws IOException, InterruptedException {
 
         List<String> command = buildAutodetectCommand();
 

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/job/process/autodetect/AutodetectCommunicator.java

@@ -196,7 +196,7 @@ public void killProcess(boolean awaitCompletion, boolean finish, boolean finaliz
             processKilled = true;
             autodetectResultProcessor.setProcessKilled();
             autodetectWorkerExecutor.shutdown();
-            autodetectProcess.kill();
+            autodetectProcess.kill(awaitCompletion);
 
             if (awaitCompletion) {
                 try {

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/job/process/autodetect/BlackHoleAutodetectProcess.java

@@ -134,7 +134,7 @@ public void close() {
     }
 
     @Override
-    public void kill() {
+    public void kill(boolean awaitCompletion) {
         open = false;
     }
 

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/job/process/autodetect/NativeAutodetectProcessFactory.java

@@ -127,6 +127,9 @@ void createNativeProcess(Job job, AutodetectParams autodetectParams, ProcessPipe
                 autodetectBuilder.quantiles(autodetectParams.quantiles());
             }
             autodetectBuilder.build();
+        } catch (InterruptedException e) {
+            Thread.currentThread().interrupt();
+            LOGGER.warn("Interrupted while launching autodetect for job " + job.getId());
         } catch (IOException e) {
             String msg = "Failed to launch autodetect for job " + job.getId();
             LOGGER.error(msg);

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/job/process/normalizer/MultiplyingNormalizerProcess.java

@@ -90,7 +90,7 @@ public void flushStream() {
     }
 
     @Override
-    public void kill() {
+    public void kill(boolean awaitCompletion) {
         // Nothing to do
     }
 

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/job/process/normalizer/NativeNormalizerProcessFactory.java

@@ -81,6 +81,9 @@ private void createNativeProcess(String jobId, String quantilesState, ProcessPip
             List<String> command = new NormalizerBuilder(env, jobId, quantilesState, bucketSpan).build();
             processPipes.addArgs(command);
             nativeController.startProcess(command);
+        } catch (InterruptedException e) {
+            Thread.currentThread().interrupt();
+            LOGGER.warn("Interrupted while launching normalizer for job " + jobId);
         } catch (IOException e) {
             String msg = "Failed to launch normalizer for job " + jobId;
             LOGGER.error(msg);

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/process/AbstractNativeProcess.java

@@ -206,19 +206,22 @@ public void close() throws IOException {
     }
 
     @Override
-    public void kill() throws IOException {
+    public void kill(boolean awaitCompletion) throws IOException {
         LOGGER.debug("[{}] Killing {} process", jobId, getName());
         processKilled = true;
         try {
             // The PID comes via the processes log stream. We do wait here to give the process the time to start up and report its PID.
             // Without the PID we cannot kill the process.
-            nativeController.killProcess(cppLogHandler().getPid(processPipes.getTimeout()));
+            nativeController.killProcess(cppLogHandler().getPid(processPipes.getTimeout()), awaitCompletion);
 
             // Wait for the process to die before closing processInStream as if the process
             // is still alive when processInStream is closed it may start persisting state
             cppLogHandler().waitForLogStreamClose(WAIT_FOR_KILL_TIMEOUT);
         } catch (TimeoutException e) {
             LOGGER.warn("[{}] Failed to get PID of {} process to kill", jobId, getName());
+        } catch (InterruptedException e) {
+            Thread.currentThread().interrupt();
+            LOGGER.warn("[{}] Interrupted while killing {} process", jobId, getName());
         } finally {
             try {
                 if (processInStream() != null) {

x-pack/plugin/ml/src/main/java/org/elasticsearch/xpack/ml/process/ControllerResponse.java

@@ -0,0 +1,86 @@
+/*
+ * 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.ml.process;
+
+import org.elasticsearch.common.ParseField;
+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;
+
+public class ControllerResponse implements ToXContentObject {
+
+    public static final ParseField TYPE = new ParseField("controller_response");
+
+    public static final ParseField COMMAND_ID = new ParseField("id");
+    public static final ParseField SUCCESS = new ParseField("success");
+    public static final ParseField REASON = new ParseField("reason");
+
+    public static final ConstructingObjectParser<ControllerResponse, Void> PARSER = new ConstructingObjectParser<>(
+        TYPE.getPreferredName(), a -> new ControllerResponse((int) a[0], (boolean) a[1], (String) a[2]));
+
+    static {
+        PARSER.declareInt(ConstructingObjectParser.constructorArg(), COMMAND_ID);
+        PARSER.declareBoolean(ConstructingObjectParser.constructorArg(), SUCCESS);
+        PARSER.declareString(ConstructingObjectParser.optionalConstructorArg(), REASON);
+    }
+
+    private final int commandId;
+    private final boolean success;
+    private final String reason;
+
+    ControllerResponse(int commandId, boolean success, String reason) {
+        this.commandId = commandId;
+        this.success = success;
+        this.reason = reason;
+    }
+
+    public int getCommandId() {
+        return commandId;
+    }
+
+    public boolean isSuccess() {
+        return success;
+    }
+
+    public String getReason() {
+        return reason;
+    }
+
+    @Override
+    public XContentBuilder toXContent(XContentBuilder builder, Params params) throws IOException {
+        builder.startObject();
+        builder.field(COMMAND_ID.getPreferredName(), commandId);
+        builder.field(SUCCESS.getPreferredName(), success);
+        if (reason != null) {
+            builder.field(REASON.getPreferredName(), reason);
+        }
+        builder.endObject();
+        return builder;
+    }
+
+    @Override
+    public boolean equals(Object o) {
+        if (this == o) {
+            return true;
+        }
+        if (o == null || getClass() != o.getClass()) {
+            return false;
+        }
+        ControllerResponse that = (ControllerResponse) o;
+        return this.commandId == that.commandId &&
+            this.success == that.success &&
+            Objects.equals(this.reason, that.reason);
+    }
+
+    @Override
+    public int hashCode() {
+        return Objects.hash(commandId, success, reason);
+    }
+}