Swap our metrics provider from statsd to otel-metrics + instrument our Fetcher threadpool

[piyush-zlai]

Summary

PR to swap our metrics reporter from statsd to open telemetry metrics. We need otel to allow us to capture metrics in Etsy without the need of a prometheus statsd exporter sidecar that they've seen issues with occasionally. Otel in general is a popular metrics ingestion interface with a number of supported backends (e.g. prom / datadog / gcloud / aws cloudwatch). Wiring up Otel also enables us to set up traces and spans in the repo in the future.
Broad changes:

• Decouple bulk of the metrics reporting logic from the Metrics.Context. The metrics reporter we use is pluggable. Currently this is just the OpenTelemetry but in principle we can support others in the future.
• Online module creates the appropriate otel SDK - either we use the Http provider or the Prometheus Http server. We need the Http provider to plug into Vert.x as their Micrometer integration works with that. The Prom http server is what Etsy is keen we use.

Checklist

• Added Unit Tests
• Covered by existing CI
• Integration tested
• Documentation update

Tested via docker container and a local instance of open telemetry:
Start up fetcher docker svc

docker run -v ~/.config/gcloud/application_default_credentials.json:/gcp/credentials.json  -p 9000:9000  -e "GCP_PROJECT_ID=canary-443022"  -e "GOOGLE_CLOUD_PROJECT=canary-443022"  -e "GCP_BIGTABLE_INSTANCE_ID=zipline-canary-instance"  -e "EXPORTER_OTLP_ENDPOINT=http://host.docker.internal:4318"  -e GOOGLE_APPLICATION_CREDENTIALS=/gcp/credentials.json  zipline-fetcher:latest

And then otel:

./otelcol --config otel-collector-config.yaml
...

We see:

2025-04-18T17:35:37.351-0400	info	ResourceMetrics #0
Resource SchemaURL: 
Resource attributes:
     -> service.name: Str(ai.chronon)
     -> telemetry.sdk.language: Str(java)
     -> telemetry.sdk.name: Str(opentelemetry)
     -> telemetry.sdk.version: Str(1.49.0)
ScopeMetrics #0
ScopeMetrics SchemaURL: 
InstrumentationScope ai.chronon 3.7.0-M11
Metric #0
Descriptor:
     -> Name: kv_store.bigtable.cache.insert
     -> Description: 
     -> Unit: 
     -> DataType: Sum
     -> IsMonotonic: true
     -> AggregationTemporality: Cumulative
NumberDataPoints #0
Data point attributes:
     -> dataset: Str(TableId{tableId=CHRONON_METADATA})
     -> environment: Str(kv_store)
     -> production: Str(false)
StartTimestamp: 2025-04-18 21:31:52.180857637 +0000 UTC
Timestamp: 2025-04-18 21:35:37.18442138 +0000 UTC
Value: 1
Metric #1
Descriptor:
     -> Name: kv_store.bigtable.multiGet.latency
     -> Description: 
     -> Unit: 
     -> DataType: Histogram
     -> AggregationTemporality: Cumulative
HistogramDataPoints #0
Data point attributes:
     -> dataset: Str(TableId{tableId=CHRONON_METADATA})
     -> environment: Str(kv_store)
     -> production: Str(false)
StartTimestamp: 2025-04-18 21:31:52.180857637 +0000 UTC
Timestamp: 2025-04-18 21:35:37.18442138 +0000 UTC
Count: 1
Sum: 229.000000
Min: 229.000000
Max: 229.000000
ExplicitBounds #0: 0.000000
...
Buckets #0, Count: 0
...

Summary by CodeRabbit

Summary by CodeRabbit

• New Features

  • Introduced OpenTelemetry-based metrics reporting, replacing previous StatsD integration.
  • Added support for OTLP and Prometheus metrics exporters with configurable endpoints.
  • Enhanced thread pool executors with built-in metrics instrumentation and detailed task timing.

• Refactor

  • Updated metrics APIs to use structured tags and a pluggable reporter interface.
  • Improved per-dataset metrics scoping and context management.

• Chores

  • Upgraded and expanded dependencies to include OpenTelemetry, Micrometer OTLP, and Prometheus libraries.
  • Updated build and environment configuration to support new metrics exporters.

[coderabbitai]

Actionable comments posted: 5

🧹 Nitpick comments (3)

service_commons/src/main/java/ai/chronon/service/ChrononServiceLauncher.java (2)

27-33: Update class comment.

Comment on lines 28-30 still refers to StatsD, should be updated to reflect OpenTelemetry usage.

---

92-94: Verify duration format.

Ensure exportInterval is in ISO-8601 format (e.g., "PT15S") for Duration.parse().

cloud_gcp/src/main/scala/ai/chronon/integrations/cloud_gcp/BigTableKVStoreImpl.scala (1)

155-159: Repeated copy(...) per call is GC‑heavy

Creating a new Context for every dataset on every call allocates objects and populates the tag cache again.
Consider caching only once per dataset and re‑using the same instance.

📜 Review details

Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro (Legacy)

📥 Commits

Reviewing files that changed from the base of the PR and between 289e64f and 1e571bd.

📒 Files selected for processing (15)

• cloud_aws/src/main/scala/ai/chronon/integrations/aws/DynamoDBKVStoreImpl.scala (2 hunks)
• cloud_gcp/src/main/scala/ai/chronon/integrations/cloud_gcp/BigTableKVStoreImpl.scala (18 hunks)
• docker/fetcher/start.sh (1 hunks)
• online/src/main/java/ai/chronon/online/JavaFetcher.java (7 hunks)
• online/src/main/scala/ai/chronon/online/Api.scala (3 hunks)
• online/src/main/scala/ai/chronon/online/fetcher/FetchContext.scala (2 hunks)
• online/src/main/scala/ai/chronon/online/fetcher/Fetcher.scala (9 hunks)
• online/src/main/scala/ai/chronon/online/fetcher/FetcherCache.scala (3 hunks)
• online/src/main/scala/ai/chronon/online/metrics/Metrics.scala (6 hunks)
• online/src/main/scala/ai/chronon/online/metrics/MetricsReporter.scala (1 hunks)
• online/src/main/scala/ai/chronon/online/metrics/OtelMetricsReporter.scala (1 hunks)
• online/src/test/scala/ai/chronon/online/test/TagsTest.scala (2 hunks)
• service/src/main/java/ai/chronon/service/FetcherVerticle.java (3 hunks)
• service_commons/src/main/java/ai/chronon/service/ChrononServiceLauncher.java (3 hunks)
• service_commons/src/main/java/ai/chronon/service/OpenTelemetryProvider.java (1 hunks)

🚧 Files skipped from review as they are similar to previous changes (8)

• online/src/main/scala/ai/chronon/online/fetcher/FetchContext.scala
• online/src/test/scala/ai/chronon/online/test/TagsTest.scala
• service/src/main/java/ai/chronon/service/FetcherVerticle.java
• cloud_aws/src/main/scala/ai/chronon/integrations/aws/DynamoDBKVStoreImpl.scala
• online/src/main/scala/ai/chronon/online/fetcher/FetcherCache.scala
• online/src/main/scala/ai/chronon/online/Api.scala
• online/src/main/scala/ai/chronon/online/fetcher/Fetcher.scala
• online/src/main/java/ai/chronon/online/JavaFetcher.java

🧰 Additional context used

🧬 Code Graph Analysis (1)

online/src/main/scala/ai/chronon/online/metrics/OtelMetricsReporter.scala (1)

online/src/main/scala/ai/chronon/online/metrics/Metrics.scala (9)

• Metrics (25-219)
• Context (89-132)
• Context (134-218)
• toTags (158-183)
• count (213-213)
• gauge (215-215)
• gauge (217-217)
• distribution (207-210)
• distribution (210-213)

⏰ Context from checks skipped due to timeout of 90000ms (35)

• GitHub Check: service_commons_tests
• GitHub Check: service_tests
• GitHub Check: cloud_gcp_tests
• GitHub Check: online_tests
• GitHub Check: cloud_aws_tests
• GitHub Check: streaming_tests
• GitHub Check: hub_tests
• GitHub Check: service_tests
• GitHub Check: join_tests
• GitHub Check: streaming_tests
• GitHub Check: flink_tests
• GitHub Check: online_tests
• GitHub Check: spark_tests
• GitHub Check: cloud_gcp_tests
• GitHub Check: groupby_tests
• GitHub Check: join_tests
• GitHub Check: hub_tests
• GitHub Check: analyzer_tests
• GitHub Check: cloud_aws_tests
• GitHub Check: groupby_tests
• GitHub Check: spark_tests
• GitHub Check: flink_tests
• GitHub Check: api_tests
• GitHub Check: fetcher_tests
• GitHub Check: batch_tests
• GitHub Check: aggregator_tests
• GitHub Check: aggregator_tests
• GitHub Check: analyzer_tests
• GitHub Check: orchestration_tests
• GitHub Check: fetcher_tests
• GitHub Check: api_tests
• GitHub Check: batch_tests
• GitHub Check: scala_compile_fmt_fix
• GitHub Check: orchestration_tests
• GitHub Check: enforce_triggered_workflows

🔇 Additional comments (6)

service_commons/src/main/java/ai/chronon/service/OpenTelemetryProvider.java (1)

11-21: Implementation looks clean.

Thread-safe singleton design with Optional for safe instance access.

service_commons/src/main/java/ai/chronon/service/ChrononServiceLauncher.java (1)

65-107: Looks good, but consider caching the OpenTelemetry instance.

The implementation correctly sets up OpenTelemetry for metrics reporting.

docker/fetcher/start.sh (1)

21-22: Log message update is appropriate.

Script correctly checks for OTLP endpoint instead of StatsD host.

online/src/main/scala/ai/chronon/online/metrics/MetricsReporter.scala (2)

5-10: Clean provider interface.

Generic type parameter enables flexibility for different client implementations.

---

12-24: Well-designed reporter trait.

Interface supports all common metric types with consistent method signatures.

online/src/main/scala/ai/chronon/online/metrics/OtelMetricsReporter.scala (1)

62-66: DoubleGauge build call may not compile on latest OTel

meter.gaugeBuilder(metric).build() returns an observable builder that expects a callback, not a direct set.
Verify against the OTel API in build.sbt; you may need ofDoubles().buildWithCallback(...) or switch to DoubleUpDownCounter.

[coderabbitai]

💡 Verification agent

🧩 Analysis chain

Verify Micrometer OTLP config keys.

Modern Micrometer versions (≥1.12) may expect different configuration keys. Consider confirming:

-                case "otlp.url":
+                case "otlp.metrics.export.url":

---

🌐 Web query:

What configuration keys does OtlpMeterRegistry use in Micrometer version 1.12+?

💡 Result:

Configuration Keys for OtlpMeterRegistry in Micrometer 1.12+

OtlpMeterRegistry uses a set of configuration keys to control behaviors such as endpoint selection, timeouts, resource attributes, and other telemetry settings. These keys are provided to the registry via the OtlpConfig interface, which supports property-based configuration—commonly via Spring Boot properties with the prefix management.otlp.metrics.export[1].

Common Configuration Keys

Below are the primary configuration keys available for OtlpMeterRegistry in Micrometer 1.12 and newer:

Key	Purpose	Default/Example Value
enabled	Whether OTLP export is enabled	true
endpoint	The OTLP metrics endpoint (URL)	(none, must be set)
step	Frequency to report metrics (duration)	PT1M (1 minute)
compression	Compression type for payload (e.g., gzip)	none
headers	Additional headers for requests (key-value pairs)	(none)
aggregation-temporality	Preferred temporality for cumulative or delta data	cumulative or delta
resource-attributes	Resource attributes attached to all metrics (comma-separated key-value pairs)	See below
timeout	Request timeout (duration)	10s
base-time-unit	Base time unit for timers	milliseconds

Resource Attributes:

• The following attributes are set by default:
  • telemetry.sdk.name: io.micrometer
  • telemetry.sdk.language: java
  • telemetry.sdk.version: Micrometer version (e.g., 1.12.0)
  • service.name: unknown_service (overridable by OTEL_SERVICE_NAME environment variable)
• Custom resource attributes can be set using the resource-attributes key, or, if empty, loaded from the OTEL_RESOURCE_ATTRIBUTES environment variable[1].

Example: Spring Boot Property Mapping

When using Spring Boot, you can set these in your application.properties or application.yaml:

management.otlp.metrics.export.enabled=true
management.otlp.metrics.export.endpoint=https://otel-collector:4318/v1/metrics
management.otlp.metrics.export.step=30s
management.otlp.metrics.export.headers=Authorization=Bearer <token>
management.otlp.metrics.export.aggregation-temporality=delta
management.otlp.metrics.export.resource-attributes=service.name=my-service,env=prod
management.otlp.metrics.export.timeout=5s

Notes

• The full set of supported keys is defined by the OtlpConfig interface. When implementing or overriding its get(String key) method, properties are fetched by their keys, typically with the prefix management.otlp.metrics.export when using Spring Boot’s auto-configuration[1][8].
• Environmental variables such as OTEL_RESOURCE_ATTRIBUTES and OTEL_SERVICE_NAME can influence resource attributes and take precedence over some configuration keys[1].
• The list above covers the most commonly used and documented keys as of Micrometer 1.12+; newer versions may introduce additional or changed keys.

For the latest list and advanced options, consult the official Micrometer OTLP documentation[1].

Citations:

• 1: https://docs.micrometer.io/micrometer/reference/implementations/otlp.html
• 2: https://github.com/micrometer-metrics/micrometer/wiki/1.13-Migration-Guide
• 3: https://docs.micrometer.io/micrometer/reference/implementations/prometheus.html
• 4: https://github.com/micrometer-metrics/micrometer/blob/main/implementations/micrometer-registry-otlp/src/main/java/io/micrometer/registry/otlp/OtlpMeterRegistry.java
• 5: https://docs.dynatrace.com/docs/ingest-from/extend-dynatrace/extend-metrics/ingestion-methods/micrometer
• 6: https://spring.io/blog/2018/03/16/micrometer-spring-boot-2-s-new-application-metrics-collector
• 7: https://docs.spring.io/spring-boot/reference/actuator/metrics.html
• 8: https://javadoc.io/static/io.micrometer/micrometer-registry-otlp/1.12.12/allclasses-index.html
• 9: https://www.javadoc.io/doc/io.micrometer/micrometer-core/1.0.3/io/micrometer/core/instrument/MeterRegistry.Config.html
• 10: https://mvnrepository.com/artifact/io.micrometer/micrometer-registry-otlp/1.12.5

---

Use correct Micrometer OTLP config keys

Micrometer 1.12+ expects the keys endpoint, step and resource-attributes, not otlp.*.

• File: service_commons/src/main/java/ai/chronon/service/ChrononServiceLauncher.java (lines 52–63)

Suggested diff:

-                case "otlp.url":
+                case "endpoint":
                     return exporterUrl;
-                case "otlp.step":
+                case "step":
                     return exportInterval;
-                case "otlp.resourceAttributes":
-                    return "service.name=" + serviceName;
+                case "resource-attributes":
+                    return "service.name=" + serviceName;

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	OtlpConfig otlpConfig = key -> {
	switch (key) {
	case "otlp.url":
	return exporterUrl;
	case "otlp.step":
	return exportInterval;
	case "otlp.resourceAttributes":
	return "service.name=" + serviceName;
	default:
	return null;
	}
	};
	OtlpConfig otlpConfig = key -> {
	switch (key) {
	case "endpoint":
	return exporterUrl;
	case "step":
	return exportInterval;
	case "resource-attributes":
	return "service.name=" + serviceName;
	default:
	return null;
	}
	};

[coderabbitai]

🛠️ Refactor suggestion

Unbounded tableToContext risks memory leak

TrieMap keeps a distinct Context per BigTable id forever. If the service touches many ephemeral tables, the map will grow without limit.
Add TTL/LRU eviction or at least expose a cleanup hook.

[piyush-zlai]

we are capped here as we have only 4 tables to contend with

[coderabbitai]

@piyush-zlai Understood! Since you're only working with 4 tables total, the memory usage for tableToContext will remain bounded and small. That's a perfectly reasonable approach given your specific use case.

---

✏️ Learnings added

Learnt from: piyush-zlai
PR: zipline-ai/chronon#657
File: cloud_gcp/src/main/scala/ai/chronon/integrations/cloud_gcp/BigTableKVStoreImpl.scala:93-97
Timestamp: 2025-04-21T15:10:40.802Z
Learning: The BigTableKVStoreImpl in the chronon codebase only interacts with 4 BigTable tables total, so unbounded caching in tableToContext is not a concern.

Note: Learnings are effective only in the context of similar code segments. To apply general coding guidelines, please configure review instructions. You can manage existing learnings in the app.

[coderabbitai]

⚠️ Potential issue

Wrong BuildInfo reference

org.json4s.BuildInfo.version is the json4s library version, not Chronon’s.
Use your project’s generated ai.chronon.BuildInfo (or similar) to avoid misleading instrumentation metadata.

[piyush-zlai]

removed this for now. Will circle back and add a version later once we wire this into our bazel setup

[coderabbitai]

@piyush-zlai That's a sensible approach. Removing the potentially incorrect BuildInfo reference now and adding the correct version information later when you have proper Bazel integration will avoid any misleading instrumentation metadata. Looking forward to seeing the final implementation when you circle back to it.

[coderabbitai]

🛠️ Refactor suggestion

Instrument caches lack eviction

counters, gauges, and histograms grow with every new metric name.
If metric names are dynamic (e.g., per‑dataset), memory will balloon.
Consider bounded caches or pre‑register a fixed set of names.

[piyush-zlai]

the names are fixed based on what we've explicitly configured in the code.

[coderabbitai]

@piyush-zlai Got it, thanks for clarifying. Since the metric names are fixed and explicitly configured in the code rather than dynamically generated, the TrieMaps won't grow unbounded and the current implementation without eviction is appropriate.

---

✏️ Learnings added

Learnt from: piyush-zlai
PR: zipline-ai/chronon#657
File: online/src/main/scala/ai/chronon/online/metrics/OtelMetricsReporter.scala:30-34
Timestamp: 2025-04-21T15:12:30.670Z
Learning: In the OtelMetricsReporter class, the instrument caches (counters, gauges, histograms) don't need eviction mechanisms because metric names are fixed and explicitly configured in the code, not dynamically generated.

Note: Learnings are effective only in the context of similar code segments. To apply general coding guidelines, please configure review instructions. You can manage existing learnings in the app.

[coderabbitai]

🛠️ Refactor suggestion

Only last join tag survives

addTag(Tag.Join, _) inside a loop overwrites previous values; multi‑join contexts lose information.
Aggregating (joinNames.mkString(",")) before a single addTag preserves all joins while keeping one key.

[coderabbitai]

Actionable comments posted: 0

♻️ Duplicate comments (5)

online/src/main/scala/ai/chronon/online/metrics/Metrics.scala (5)

205-236: ⚠️ Potential issue

toTags never consumed – metrics lose all context

All reporter methods use empty maps instead of the context's tags.

-    def increment(metric: String): Unit = Context.client.count(prefix(metric), 1, Map.empty)
+    def increment(metric: String): Unit = Context.client.count(prefix(metric), 1, toTags)

-    def distribution(metric: String, value: Long): Unit =
-      Context.client.distribution(prefix(metric), value, Map.empty)
+    def distribution(metric: String, value: Long): Unit =
+      Context.client.distribution(prefix(metric), value, toTags)

-    def count(metric: String, value: Long): Unit = Context.client.count(prefix(metric), value)
+    def count(metric: String, value: Long): Unit = Context.client.count(prefix(metric), value, toTags)

-    def gauge(metric: String, value: Long): Unit = Context.client.longGauge(prefix(metric), value)
+    def gauge(metric: String, value: Long): Unit = Context.client.longGauge(prefix(metric), value, toTags)

-    def gauge(metric: String, value: Double): Unit = Context.client.doubleGauge(prefix(metric), value)
+    def gauge(metric: String, value: Double): Unit = Context.client.doubleGauge(prefix(metric), value, toTags)

---

176-201: ⚠️ Potential issue

Only last join tag survives

addTag(Tag.Join, _) inside a loop overwrites previous values; multi‑join contexts lose information.

Fix by aggregating join names before a single addTag call:

-      joinNames.foreach(addTag(Tag.Join, _))
+      if (joinNames.nonEmpty) {
+        addTag(Tag.Join, joinNames.mkString(","))
+      }

---

222-222: ⚠️ Potential issue

Merge with context tags

Exception tags should merge with context tags, not replace them.

-      Context.client.count(prefix(Name.Exception), 1, Map(Metrics.Name.Exception -> exceptionSignature))
+      Context.client.count(prefix(Name.Exception), 1, toTags + (Metrics.Name.Exception -> exceptionSignature))

---

228-229: ⚠️ Potential issue

Merge with context tags

Additional tags should merge with context tags, not replace them.

-    def distribution(metric: String, value: Long, additionalTags: Map[String, String]): Unit =
-      Context.client.distribution(prefix(metric), value, additionalTags)
+    def distribution(metric: String, value: Long, additionalTags: Map[String, String]): Unit =
+      Context.client.distribution(prefix(metric), value, toTags ++ additionalTags)

---

207-208: ⚠️ Potential issue

Merge with context tags

Additional tags should merge with context tags, not replace them.

-    def increment(metric: String, additionalTags: Map[String, String]): Unit =
-      Context.client.count(prefix(metric), 1, additionalTags)
+    def increment(metric: String, additionalTags: Map[String, String]): Unit =
+      Context.client.count(prefix(metric), 1, toTags ++ additionalTags)

🧹 Nitpick comments (3)

online/src/main/scala/ai/chronon/online/metrics/FlexibleExecutionContext.scala (1)

38-44: Queue capacity increased by 10x

The queue capacity was significantly increased from 1000 to 10000 items. While this prevents task rejection under high load, monitor memory usage.

online/src/main/scala/ai/chronon/online/metrics/InstrumentedThreadPoolExecutor.scala (1)

30-30: Consider constructor-injected metrics context

Fixed context limits reusability. Consider accepting it as a constructor parameter.

-class InstrumentedThreadPoolExecutor(corePoolSize: Int,
+class InstrumentedThreadPoolExecutor(corePoolSize: Int,
                                     maximumPoolSize: Int,
                                     keepAliveTime: Long,
                                     unit: TimeUnit,
                                     workQueue: BlockingQueue[Runnable],
                                     threadFactory: ThreadFactory,
+                                    metricsContext: Metrics.Context = Metrics.Context(Metrics.Environment.Fetcher).withSuffix("threadpool"),
                                     metricsIntervalSeconds: Int = 15)
   extends ThreadPoolExecutor(
     corePoolSize,
     maximumPoolSize,
     keepAliveTime,
     unit,
     workQueue,
     threadFactory
   ) {
 // Reporter for periodic metrics
 private val metricsReporter: ScheduledExecutorService = buildMetricsScheduledExecutor()

-protected val metricsContext: Metrics.Context = Metrics.Context(Metrics.Environment.Fetcher).withSuffix("threadpool")

online/src/main/scala/ai/chronon/online/metrics/Metrics.scala (1)

130-146: Add error handling for OtelMetricsReporter creation

Current implementation will throw exceptions on initialization failures. Consider adding fallback.

   private val client: MetricsReporter = {
     // Can disable metrics collection for local / dev environments
     val metricsEnabled: Boolean = System.getProperty(MetricsEnabled, "true").toBoolean
     val reporter: String = System.getProperty(MetricsReporter, "otel")

-    reporter.toLowerCase match {
-      case "otel" | "opentelemetry" =>
-        if (metricsEnabled) {
-          val metricReader = OtelMetricsReporter.buildOtelMetricReader()
-          val openTelemetry = OtelMetricsReporter.buildOpenTelemetryClient(metricReader)
-          new OtelMetricsReporter(openTelemetry)
-        } else {
-          new OtelMetricsReporter(OpenTelemetry.noop())
-        }
-      case _ =>
-        throw new IllegalArgumentException(s"Unknown metrics reporter: $reporter. Only opentelemetry is supported.")
+    try {
+      reporter.toLowerCase match {
+        case "otel" | "opentelemetry" =>
+          if (metricsEnabled) {
+            val metricReader = OtelMetricsReporter.buildOtelMetricReader()
+            val openTelemetry = OtelMetricsReporter.buildOpenTelemetryClient(metricReader)
+            new OtelMetricsReporter(openTelemetry)
+          } else {
+            new OtelMetricsReporter(OpenTelemetry.noop())
+          }
+        case _ =>
+          throw new IllegalArgumentException(s"Unknown metrics reporter: $reporter. Only opentelemetry is supported.")
+      }
+    } catch {
+      case e: Exception =>
+        org.slf4j.LoggerFactory.getLogger(getClass).error(s"Failed to initialize metrics: ${e.getMessage}", e)
+        new OtelMetricsReporter(OpenTelemetry.noop())
+    }

📜 Review details

Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro (Legacy)

📥 Commits

Reviewing files that changed from the base of the PR and between 9ef2fa9 and 54c82bd.

📒 Files selected for processing (15)

• cloud_gcp/src/main/scala/ai/chronon/integrations/cloud_gcp/BigTableKVStoreImpl.scala (17 hunks)
• docker/fetcher/start.sh (2 hunks)
• maven_install.json (37 hunks)
• online/BUILD.bazel (4 hunks)
• online/src/main/java/ai/chronon/online/JavaFetcher.java (1 hunks)
• online/src/main/scala/ai/chronon/online/fetcher/Fetcher.scala (2 hunks)
• online/src/main/scala/ai/chronon/online/metrics/FlexibleExecutionContext.scala (1 hunks)
• online/src/main/scala/ai/chronon/online/metrics/InstrumentedThreadPoolExecutor.scala (1 hunks)
• online/src/main/scala/ai/chronon/online/metrics/Metrics.scala (6 hunks)
• online/src/main/scala/ai/chronon/online/metrics/MetricsReporter.scala (1 hunks)
• online/src/main/scala/ai/chronon/online/metrics/OtelMetricsReporter.scala (1 hunks)
• online/src/test/scala/ai/chronon/online/test/TagsTest.scala (2 hunks)
• service_commons/BUILD.bazel (1 hunks)
• service_commons/src/main/java/ai/chronon/service/ChrononServiceLauncher.java (1 hunks)
• tools/build_rules/dependencies/maven_repository.bzl (2 hunks)

✅ Files skipped from review due to trivial changes (2)

• online/src/main/java/ai/chronon/online/JavaFetcher.java
• online/src/main/scala/ai/chronon/online/fetcher/Fetcher.scala

🚧 Files skipped from review as they are similar to previous changes (8)

• service_commons/BUILD.bazel
• online/src/test/scala/ai/chronon/online/test/TagsTest.scala
• online/BUILD.bazel
• docker/fetcher/start.sh
• service_commons/src/main/java/ai/chronon/service/ChrononServiceLauncher.java
• tools/build_rules/dependencies/maven_repository.bzl
• cloud_gcp/src/main/scala/ai/chronon/integrations/cloud_gcp/BigTableKVStoreImpl.scala
• online/src/main/scala/ai/chronon/online/metrics/OtelMetricsReporter.scala

🧰 Additional context used

🧬 Code Graph Analysis (1)

online/src/main/scala/ai/chronon/online/metrics/InstrumentedThreadPoolExecutor.scala (2)

online/src/main/scala/ai/chronon/online/metrics/Metrics.scala (3)

• Metrics (26-237)
• Environment (27-43)
• withSuffix (165-165)

online/src/main/scala/ai/chronon/online/metrics/OtelMetricsReporter.scala (1)

• distribution (73-78)

⏰ Context from checks skipped due to timeout of 90000ms (21)

• GitHub Check: cloud_aws_tests
• GitHub Check: service_tests
• GitHub Check: service_tests
• GitHub Check: online_tests
• GitHub Check: service_commons_tests
• GitHub Check: hub_tests
• GitHub Check: hub_tests
• GitHub Check: online_tests
• GitHub Check: flink_tests
• GitHub Check: cloud_gcp_tests
• GitHub Check: cloud_gcp_tests
• GitHub Check: cloud_aws_tests
• GitHub Check: aggregator_tests
• GitHub Check: flink_tests
• GitHub Check: api_tests
• GitHub Check: aggregator_tests
• GitHub Check: orchestration_tests
• GitHub Check: orchestration_tests
• GitHub Check: scala_compile_fmt_fix
• GitHub Check: api_tests
• GitHub Check: enforce_triggered_workflows

🔇 Additional comments (40)

online/src/main/scala/ai/chronon/online/metrics/MetricsReporter.scala (1)

8-17: LGTM - Well-structured metrics interface

Clean, consistent method signatures with appropriate parameter defaults.

online/src/main/scala/ai/chronon/online/metrics/InstrumentedThreadPoolExecutor.scala (2)

66-89: LGTM - Well-implemented task instrumentation

Properly captures and reports wait time, execution time, and total time metrics.

---

91-100: LGTM - Proper resource cleanup

Correctly shuts down metrics reporter along with thread pool.

maven_install.json (37)

3-4: Metadata bump

---

1197-1203: OkHttp artifact added

---

1778-1784: Micrometer OTLP registry added

---

2131-2137: OpenTelemetry proto added

---

2147-2150: OTel API version bumped

---

2159-2249: OTel SDK modules updated to 1.49.0

---

2257-2298: Prometheus metrics modules added

---

4172-4185: Kotlin stdlib JDK7/JDK8 added

---

6420-6426: OkHttp/Okio dependency mapping

---

6738-6741: Micrometer OTLP mapping

---

7007-7024: OTel dependency mappings updated

---

8476-8482: Kotlin stdlib mapping updated

---

11312-11329: OkHttp package-to-class mapping

---

12166-12168: Micrometer OTLP package mapping

---

12540-12549: OTel proto package mapping

---

12578-12615: OTel exporter package mappings

---

12628-12631: SDK autoconfigure SPI mapping

---

12678-12699: Prometheus package mappings

---

24457-24463: Sources list includes OkHttp/Okio

---

24623-24624: Micrometer OTLP in sources

---

24729-24731: OTel proto in sources

---

24739-24747: OTel exporter in sources

---

24765-24775: Prometheus in sources

---

25298-25301: Kotlin JDK variants in sources

---

25925-25930: OkHttp/Okio in sources

---

26091-26092: Micrometer OTLP in sources

---

26197-26199: OTel proto in sources

---

26207-26215: OTel exporter in sources

---

26233-26243: Prometheus in sources

---

26766-26769: Kotlin JDK sources updated

---

27393-27398: OkHttp/Okio sources updated

---

27559-27560: Micrometer OTLP in sources

---

27665-27667: OTel proto in sources

---

27675-27683: OTel exporter sources

---

27698-27712: Prometheus sources list

---

28234-28237: Kotlin JDK in stdlib sources

---

30256-30303: SPI registrations for OTel exporters finalized