Add Telemetry

CHANGELOG.md

@@ -6,12 +6,12 @@
 
 - [API] Added `lithops.concurrent.futures`, a `concurrent.futures`-compatible executor interface backed by Lithops.
 - [Monitoring] Added Redis, AWS SQS, GCP Pub/Sub and Azure Queue Storage monitoring backends.
-- [Monitoring] Added the `aws_sqs`, `gcp_pubsub` and `azure_queue` config sections, which fall back to the `aws`, `gcp` and `azure_storage` credentials.
 - [Core] Added a cache of serialized functions to avoid re-uploading the same function.
 - [Core] Added `clean_jobs` to `wait()`, to keep the temporary data until the results are read.
 - [AWS Batch] Added the `instance_types` config option for EC2/SPOT compute environments.
 - [Multiprocessing] Added `timeout` to `acquire()`, and `_getvalue()`, `_callmethod()` and `copy_proxy()` to the manager proxies.
 - [Multiprocessing] Added `ThreadPool`, `Manager`, the standard error classes and the module-level helpers (`freeze_support()`, `get_logger()`, `log_to_stderr()`).
+- [Telemetry] Added `lithops.telemetry`, which turns the call statuses the monitor reads into 29 Prometheus or OpenTelemetry metrics.
 - [Tests] Added a unit test suite for all non-backend modules.
 
 ### Changed

config/config_template.yaml

@@ -34,7 +34,8 @@
     #log_format: "%(asctime)s [%(levelname)s] %(name)s -- %(message)s"
     #log_stream: ext://sys.stderr
     #log_filename: <PATH_TO_LOG_FILE>
-    #telemetry: False                     # Enable Prometheus push-gateway metrics
+    #telemetry: False                     # Client-side metrics. One of: False, prometheus, otlp
+    #telemetry_interval: 10               # Seconds between metric pushes. Default: 10
 
 
 # =============================================================================
@@ -584,7 +585,24 @@
 
 
 # =============================================================================
-# Prometheus – pushed metrics when `lithops.telemetry=True`
+# Prometheus – metrics pushed when `lithops.telemetry=prometheus`
 # =============================================================================
 #prometheus:
-    #apigateway: <PUSH_GATEWAY_URL>       # Prometheus Pushgateway base URL
+    #gateway: http://localhost:9091       # Pushgateway base URL. Default: http://localhost:9091
+    #job: lithops                         # Pushgateway job label. Default: lithops
+    #instance: <HOSTNAME>                 # Pushgateway instance label. Default: the hostname
+    #username: <USERNAME>                 # Basic auth user, when the gateway needs one
+    #password: <PASSWORD>                 # Basic auth password, when the gateway needs one
+    #timeout: 10                          # Push timeout in seconds. Default: 10
+
+
+# =============================================================================
+# OpenTelemetry – metrics exported when `lithops.telemetry=otlp`
+# =============================================================================
+#otlp:
+    #endpoint: http://localhost:4318      # Collector endpoint. Default: http://localhost:4318
+    #protocol: http/protobuf              # One of: http/protobuf, grpc. Default: http/protobuf
+    #service_name: lithops                # OTel service.name. Default: lithops
+    #instance: <HOSTNAME>                 # OTel host.name. Default: the hostname
+    #headers: {}                          # Extra headers, e.g. authentication
+    #timeout: 10                          # Export timeout in seconds. Default: 10

docs/index.rst

@@ -276,6 +276,7 @@ Lithops is an open-source project, actively maintained and supported by a commun
    :caption: Advanced Features
 
    source/monitoring.rst
+   source/metrics.rst
    Custom Runtimes <https://github.com/lithops-cloud/lithops/tree/master/runtime>
 
 
@@ -284,7 +285,6 @@ Lithops is an open-source project, actively maintained and supported by a commun
    :maxdepth: 0
    :caption: Experimental Features
 
-   source/metrics.rst
    source/dso.rst
 
 

docs/source/design.rst

@@ -75,6 +75,10 @@ directly to a module or class in the ``lithops`` package:
      - ``lithops/monitoring/``
      - Tracks completion through either ``StorageMonitor`` (polling) or
        ``RabbitmqMonitor`` (push notifications).
+   * - Telemetry
+     - ``lithops/telemetry/``
+     - Turns the call statuses the monitor reads into metrics, and pushes them
+       to Prometheus or to an OpenTelemetry collector. Off by default.
    * - ``InternalStorage``
      - ``lithops/storage/storage.py``
      - Uniform storage client used by both the client and the workers, backed by