[ML] More advanced model snapshot retention options (#56125)

This PR implements the following changes to make ML model snapshot
retention more flexible in advance of adding a UI for the feature in
an upcoming release.

- The default for `model_snapshot_retention_days` for new jobs is now
  10 instead of 1
- There is a new job setting, `daily_model_snapshot_retention_after_days`,
  that defaults to 1 for new jobs and `model_snapshot_retention_days`
  for pre-7.8 jobs
- For days that are older than `model_snapshot_retention_days`, all
  model snapshots are deleted as before
- For days that are in between `daily_model_snapshot_retention_after_days`
  and `model_snapshot_retention_days` all but the first model snapshot
  for that day are deleted
- The `retain` setting of model snapshots is still respected to allow
  selected model snapshots to be retained indefinitely

Closes #52150

Author: David Roberts

client/rest-high-level/src/test/java/org/elasticsearch/client/MachineLearningIT.java

@@ -2519,10 +2519,12 @@ private static Job buildJobForExpiredDataTests(String jobId) {
             .setFunction("count")
             .setDetectorDescription(randomAlphaOfLength(10))
             .build();
-        AnalysisConfig.Builder configBuilder = new AnalysisConfig.Builder(Arrays.asList(detector));
+        AnalysisConfig.Builder configBuilder = new AnalysisConfig.Builder(Collections.singletonList(detector));
         //should not be random, see:https://github.com/elastic/ml-cpp/issues/208
         configBuilder.setBucketSpan(new TimeValue(1, TimeUnit.HOURS));
         builder.setAnalysisConfig(configBuilder);
+        builder.setModelSnapshotRetentionDays(1L);
+        builder.setDailyModelSnapshotRetentionAfterDays(1L);
 
         DataDescription.Builder dataDescription = new DataDescription.Builder();
         dataDescription.setTimeFormat(DataDescription.EPOCH_MS);

client/rest-high-level/src/test/java/org/elasticsearch/client/ml/job/config/JobTests.java

@@ -141,11 +141,17 @@ public static Job.Builder createRandomizedJobBuilder() {
         if (randomBoolean()) {
             builder.setBackgroundPersistInterval(TimeValue.timeValueHours(randomIntBetween(1, 24)));
         }
+        Long modelSnapshotRetentionDays = null;
         if (randomBoolean()) {
-            builder.setModelSnapshotRetentionDays(randomNonNegativeLong());
+            modelSnapshotRetentionDays = randomNonNegativeLong();
+            builder.setModelSnapshotRetentionDays(modelSnapshotRetentionDays);
         }
         if (randomBoolean()) {
-            builder.setDailyModelSnapshotRetentionAfterDays(randomNonNegativeLong());
+            if (modelSnapshotRetentionDays != null) {
+                builder.setDailyModelSnapshotRetentionAfterDays(randomLongBetween(0, modelSnapshotRetentionDays));
+            } else {
+                builder.setDailyModelSnapshotRetentionAfterDays(randomNonNegativeLong());
+            }
         }
         if (randomBoolean()) {
             builder.setResultsRetentionDays(randomNonNegativeLong());

docs/reference/ml/anomaly-detection/apis/get-job.asciidoc

@@ -137,7 +137,8 @@ The API returns the following results:
       "model_plot_config" : {
         "enabled" : true
       },
-      "model_snapshot_retention_days" : 1,
+      "model_snapshot_retention_days" : 10,
+      "daily_model_snapshot_retention_after_days" : 1,
       "custom_settings" : {
         "created_by" : "ml-module-sample",
         ...

docs/reference/ml/anomaly-detection/apis/get-ml-info.asciidoc

@@ -102,7 +102,8 @@ This is a possible response:
       },
       "model_memory_limit" : "1gb",
       "categorization_examples_limit" : 4,
-      "model_snapshot_retention_days" : 1
+      "model_snapshot_retention_days" : 10,
+      "daily_model_snapshot_retention_after_days" : 1
     },
     "datafeeds" : {
       "scroll_size" : 1000

docs/reference/ml/anomaly-detection/apis/put-job.asciidoc

@@ -224,6 +224,10 @@ include::{docdir}/ml/ml-shared.asciidoc[tag=custom-settings]
 include::{docdir}/ml/ml-shared.asciidoc[tag=data-description]
 //End data_description
 
+`daily_model_snapshot_retention_after_days`::
+(Optional, long)
+include::{docdir}/ml/ml-shared.asciidoc[tag=daily-model-snapshot-retention-after-days]
+
 `description`::
   (Optional, string) A description of the job.
 
@@ -320,7 +324,8 @@ When the job is created, you receive the following results:
     "time_field" : "timestamp",
     "time_format" : "epoch_ms"
   },
-  "model_snapshot_retention_days" : 1,
+  "model_snapshot_retention_days" : 10,
+  "daily_model_snapshot_retention_after_days" : 1,
   "results_index_name" : "shared",
   "allow_lazy_open" : false
 }

docs/reference/ml/anomaly-detection/apis/update-job.asciidoc

@@ -82,6 +82,10 @@ close the job, then reopen the job and restart the {dfeed} for the changes to ta
 (object)
 include::{docdir}/ml/ml-shared.asciidoc[tag=custom-settings]
 
+`daily_model_snapshot_retention_after_days`::
+(long)
+include::{docdir}/ml/ml-shared.asciidoc[tag=daily-model-snapshot-retention-after-days]
+
 `description`::
 (string) A description of the job.
 

docs/reference/ml/ml-shared.asciidoc

@@ -361,6 +361,18 @@ example, it can contain custom URL information as shown in
 {ml-docs}/ml-configuring-url.html[Adding custom URLs to {ml} results].
 end::custom-settings[]
 
+tag::daily-model-snapshot-retention-after-days[]
+Advanced configuration option. Specifies a number of days between 0 and the
+value of `model_snapshot_retention_days`. After this period of time, only the first
+model snapshot per day is retained for this job. Age is calculated relative to
+the timestamp of the newest model snapshot. For new jobs, the default value is
+`1`, which means that all snapshots are retained for one day. Older snapshots
+are thinned out such that only one per day is retained. For jobs that were
+created before this setting was available, the default value matches the 
+`model_snapshot_retention_days` value, which preserves the original behavior
+and no thinning out of model snapshots occurs.
+end::daily-model-snapshot-retention-after-days[]
+
 tag::data-description[]
 The data description defines the format of the input data when you send data to
 the job by using the <<ml-post-data,post data>> API. Note that when configure
@@ -997,8 +1009,8 @@ end::model-snapshot-id[]
 tag::model-snapshot-retention-days[]
 Advanced configuration option. The period of time (in days) that model snapshots 
 are retained. Age is calculated relative to the timestamp of the newest model 
-snapshot. The default value is `1`, which means snapshots that are one day 
-(twenty-four hours) older than the newest snapshot are deleted.
+snapshot. The default value is `10`, which means snapshots that are ten days 
+older than the newest snapshot are deleted.
 end::model-snapshot-retention-days[]
 
 tag::model-timestamp[]

x-pack/plugin/core/src/main/java/org/elasticsearch/xpack/core/ml/job/config/Job.java

@@ -98,7 +98,8 @@ public class Job extends AbstractDiffable<Job> implements Writeable, ToXContentO
      */
     public static final ByteSizeValue PROCESS_MEMORY_OVERHEAD = new ByteSizeValue(10, ByteSizeUnit.MB);
 
-    public static final long DEFAULT_MODEL_SNAPSHOT_RETENTION_DAYS = 1;
+    public static final long DEFAULT_MODEL_SNAPSHOT_RETENTION_DAYS = 10;
+    public static final long DEFAULT_DAILY_MODEL_SNAPSHOT_RETENTION_AFTER_DAYS = 1;
 
     private static ObjectParser<Builder, Void> createParser(boolean ignoreUnknownFields) {
         ObjectParser<Builder, Void> parser = new ObjectParser<>("job_details", ignoreUnknownFields, Builder::new);
@@ -808,6 +809,10 @@ public Builder setModelSnapshotRetentionDays(Long modelSnapshotRetentionDays) {
             return this;
         }
 
+        public Long getModelSnapshotRetentionDays() {
+            return modelSnapshotRetentionDays;
+        }
+
         public Builder setDailyModelSnapshotRetentionAfterDays(Long dailyModelSnapshotRetentionAfterDays) {
             this.dailyModelSnapshotRetentionAfterDays = dailyModelSnapshotRetentionAfterDays;
             return this;
@@ -1043,9 +1048,6 @@ public void validateInputFields() {
 
             checkValidBackgroundPersistInterval();
             checkValueNotLessThan(0, RENORMALIZATION_WINDOW_DAYS.getPreferredName(), renormalizationWindowDays);
-            checkValueNotLessThan(0, MODEL_SNAPSHOT_RETENTION_DAYS.getPreferredName(), modelSnapshotRetentionDays);
-            checkValueNotLessThan(0, DAILY_MODEL_SNAPSHOT_RETENTION_AFTER_DAYS.getPreferredName(),
-                dailyModelSnapshotRetentionAfterDays);
             checkValueNotLessThan(0, RESULTS_RETENTION_DAYS.getPreferredName(), resultsRetentionDays);
 
             if (!MlStrings.isValidId(id)) {
@@ -1055,6 +1057,8 @@ public void validateInputFields() {
                 throw new IllegalArgumentException(Messages.getMessage(Messages.JOB_CONFIG_ID_TOO_LONG, MlStrings.ID_LENGTH_LIMIT));
             }
 
+            validateModelSnapshotRetentionSettings();
+
             validateGroups();
 
             // Results index name not specified in user input means use the default, so is acceptable in this validation
@@ -1076,6 +1080,37 @@ public void validateAnalysisLimitsAndSetDefaults(@Nullable ByteSizeValue maxMode
                     AnalysisLimits.DEFAULT_MODEL_MEMORY_LIMIT_MB);
         }
 
+        /**
+         * This is meant to be called when a new job is created.
+         * It sets {@link #dailyModelSnapshotRetentionAfterDays} to the default value if it is not set and the default makes sense.
+         */
+        public void validateModelSnapshotRetentionSettingsAndSetDefaults() {
+            validateModelSnapshotRetentionSettings();
+            if (dailyModelSnapshotRetentionAfterDays == null &&
+                modelSnapshotRetentionDays != null &&
+                modelSnapshotRetentionDays > DEFAULT_DAILY_MODEL_SNAPSHOT_RETENTION_AFTER_DAYS) {
+                dailyModelSnapshotRetentionAfterDays = DEFAULT_DAILY_MODEL_SNAPSHOT_RETENTION_AFTER_DAYS;
+            }
+        }
+
+        /**
+         * Validates that {@link #modelSnapshotRetentionDays} and {@link #dailyModelSnapshotRetentionAfterDays} make sense,
+         * both individually and in combination.
+         */
+        public void validateModelSnapshotRetentionSettings() {
+
+            checkValueNotLessThan(0, MODEL_SNAPSHOT_RETENTION_DAYS.getPreferredName(), modelSnapshotRetentionDays);
+            checkValueNotLessThan(0, DAILY_MODEL_SNAPSHOT_RETENTION_AFTER_DAYS.getPreferredName(),
+                dailyModelSnapshotRetentionAfterDays);
+
+            if (modelSnapshotRetentionDays != null &&
+                dailyModelSnapshotRetentionAfterDays != null &&
+                dailyModelSnapshotRetentionAfterDays > modelSnapshotRetentionDays) {
+                throw new IllegalArgumentException(Messages.getMessage(Messages.JOB_CONFIG_MODEL_SNAPSHOT_RETENTION_SETTINGS_INCONSISTENT,
+                    dailyModelSnapshotRetentionAfterDays, modelSnapshotRetentionDays));
+            }
+        }
+
         private void validateGroups() {
             for (String group : this.groups) {
                 if (MlStrings.isValidId(group) == false) {

x-pack/plugin/core/src/main/java/org/elasticsearch/xpack/core/ml/job/messages/Messages.java

@@ -6,6 +6,7 @@
 package org.elasticsearch.xpack.core.ml.job.messages;
 
 import org.elasticsearch.xpack.core.ml.MachineLearningField;
+import org.elasticsearch.xpack.core.ml.job.config.Job;
 
 import java.text.MessageFormat;
 import java.util.Locale;
@@ -212,6 +213,9 @@ public final class Messages {
             "This job would cause a mapping clash with existing field [{0}] - avoid the clash by assigning a dedicated results index";
     public static final String JOB_CONFIG_TIME_FIELD_NOT_ALLOWED_IN_ANALYSIS_CONFIG =
             "data_description.time_field may not be used in the analysis_config";
+    public static final String JOB_CONFIG_MODEL_SNAPSHOT_RETENTION_SETTINGS_INCONSISTENT =
+            "The value of '" + Job.DAILY_MODEL_SNAPSHOT_RETENTION_AFTER_DAYS + "' [{0}] cannot be greater than '" +
+                Job.MODEL_SNAPSHOT_RETENTION_DAYS + "' [{1}]";
 
     public static final String JOB_AND_GROUP_NAMES_MUST_BE_UNIQUE =
             "job and group names must be unique but job [{0}] and group [{0}] have the same name";

x-pack/plugin/core/src/main/java/org/elasticsearch/xpack/core/ml/job/results/ReservedFieldNames.java

@@ -231,6 +231,7 @@ public final class ReservedFieldNames {
             Job.RENORMALIZATION_WINDOW_DAYS.getPreferredName(),
             Job.BACKGROUND_PERSIST_INTERVAL.getPreferredName(),
             Job.MODEL_SNAPSHOT_RETENTION_DAYS.getPreferredName(),
+            Job.DAILY_MODEL_SNAPSHOT_RETENTION_AFTER_DAYS.getPreferredName(),
             Job.RESULTS_RETENTION_DAYS.getPreferredName(),
             Job.MODEL_SNAPSHOT_ID.getPreferredName(),
             Job.MODEL_SNAPSHOT_MIN_VERSION.getPreferredName(),