Observability
Monitor your client's activity for optimization and debugging.
redis-py has built-in support for OpenTelemetry (OTel)
instrumentation to collect metrics. This can be very helpful for
diagnosing problems and improving the performance and connection resiliency of
your application. See the
Observability overview
for an introduction to Redis client observability and a reference guide for the
available metrics.
This page explains how to enable and use OTel instrumentation
in redis-py using an example configuration for a local Grafana
instance. See our
observability demonstration repository
on GitHub to learn how to set up a suitable Grafana dashboard.
Installation
Install OTel support for redis-py with the following command:
pip install redis[otel]
Import
Start by importing the required OTel and Redis modules:
# OpenTelemetry metrics API
from opentelemetry import metrics
from opentelemetry.sdk.metrics import MeterProvider
from opentelemetry.sdk.metrics.export import PeriodicExportingMetricReader
from opentelemetry.exporter.otlp.proto.http.metric_exporter import OTLPMetricExporter
# Redis observability API
from redis.observability.providers import get_observability_instance
from redis.observability.config import OTelConfig, MetricGroup
# Redis client
import redis
exporter = OTLPMetricExporter(endpoint="http://localhost:4318/v1/metrics")
reader = PeriodicExportingMetricReader(exporter=exporter, export_interval_millis=10000)
provider = MeterProvider(metric_readers=[reader])
metrics.set_meter_provider(provider)
otel = get_observability_instance()
otel.init(OTelConfig(
# Metric groups to enable (default: CONNECTION_BASIC | RESILIENCY)
metric_groups=[
MetricGroup.CONNECTION_BASIC, # Connection creation time, relaxed timeout
MetricGroup.CONNECTION_ADVANCED, # Connection wait time, timeouts, closed connections
MetricGroup.COMMAND, # Command execution duration
MetricGroup.RESILIENCY, # Error counts, maintenance notifications
MetricGroup.PUBSUB, # PubSub message counts
MetricGroup.STREAMING, # Stream message lag
MetricGroup.CSC, # Client Side Caching metrics
],
# Filter which commands to track
include_commands=['GET', 'SET', 'HGET'], # Only track these commands
# OR
exclude_commands=['DEBUG', 'SLOWLOG'], # Track all except these
# Privacy controls
hide_pubsub_channel_names=True, # Hide channel names in PubSub metrics
hide_stream_names=True, # Hide stream names in streaming metrics
# Custom histogram buckets
buckets_operation_duration=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_stream_processing_duration=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_connection_create_time=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_connection_wait_time=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
))
r = redis.Redis(host='localhost', port=6379)
r.set('key', 'value') # Metrics collected automatically
r.get('key')
otel.shutdown()
Configure the meter provider
Otel uses a Meter provider to create the objects that collect the metric information. The example below configures a meter provider to export metrics to a local Grafana instance every 10 seconds, but see the OpenTelemetry Python docs to learn more about other export options.
# OpenTelemetry metrics API
from opentelemetry import metrics
from opentelemetry.sdk.metrics import MeterProvider
from opentelemetry.sdk.metrics.export import PeriodicExportingMetricReader
from opentelemetry.exporter.otlp.proto.http.metric_exporter import OTLPMetricExporter
# Redis observability API
from redis.observability.providers import get_observability_instance
from redis.observability.config import OTelConfig, MetricGroup
# Redis client
import redis
exporter = OTLPMetricExporter(endpoint="http://localhost:4318/v1/metrics")
reader = PeriodicExportingMetricReader(exporter=exporter, export_interval_millis=10000)
provider = MeterProvider(metric_readers=[reader])
metrics.set_meter_provider(provider)
otel = get_observability_instance()
otel.init(OTelConfig(
# Metric groups to enable (default: CONNECTION_BASIC | RESILIENCY)
metric_groups=[
MetricGroup.CONNECTION_BASIC, # Connection creation time, relaxed timeout
MetricGroup.CONNECTION_ADVANCED, # Connection wait time, timeouts, closed connections
MetricGroup.COMMAND, # Command execution duration
MetricGroup.RESILIENCY, # Error counts, maintenance notifications
MetricGroup.PUBSUB, # PubSub message counts
MetricGroup.STREAMING, # Stream message lag
MetricGroup.CSC, # Client Side Caching metrics
],
# Filter which commands to track
include_commands=['GET', 'SET', 'HGET'], # Only track these commands
# OR
exclude_commands=['DEBUG', 'SLOWLOG'], # Track all except these
# Privacy controls
hide_pubsub_channel_names=True, # Hide channel names in PubSub metrics
hide_stream_names=True, # Hide stream names in streaming metrics
# Custom histogram buckets
buckets_operation_duration=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_stream_processing_duration=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_connection_create_time=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_connection_wait_time=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
))
r = redis.Redis(host='localhost', port=6379)
r.set('key', 'value') # Metrics collected automatically
r.get('key')
otel.shutdown()
Configure the Redis client
You configure the client library for OTel only once per application. This will
enable OTel for all Redis connections you create. The example below shows the
options you can pass to the observability instance via the OTelConfig object
during initialization.
# OpenTelemetry metrics API
from opentelemetry import metrics
from opentelemetry.sdk.metrics import MeterProvider
from opentelemetry.sdk.metrics.export import PeriodicExportingMetricReader
from opentelemetry.exporter.otlp.proto.http.metric_exporter import OTLPMetricExporter
# Redis observability API
from redis.observability.providers import get_observability_instance
from redis.observability.config import OTelConfig, MetricGroup
# Redis client
import redis
exporter = OTLPMetricExporter(endpoint="http://localhost:4318/v1/metrics")
reader = PeriodicExportingMetricReader(exporter=exporter, export_interval_millis=10000)
provider = MeterProvider(metric_readers=[reader])
metrics.set_meter_provider(provider)
otel = get_observability_instance()
otel.init(OTelConfig(
# Metric groups to enable (default: CONNECTION_BASIC | RESILIENCY)
metric_groups=[
MetricGroup.CONNECTION_BASIC, # Connection creation time, relaxed timeout
MetricGroup.CONNECTION_ADVANCED, # Connection wait time, timeouts, closed connections
MetricGroup.COMMAND, # Command execution duration
MetricGroup.RESILIENCY, # Error counts, maintenance notifications
MetricGroup.PUBSUB, # PubSub message counts
MetricGroup.STREAMING, # Stream message lag
MetricGroup.CSC, # Client Side Caching metrics
],
# Filter which commands to track
include_commands=['GET', 'SET', 'HGET'], # Only track these commands
# OR
exclude_commands=['DEBUG', 'SLOWLOG'], # Track all except these
# Privacy controls
hide_pubsub_channel_names=True, # Hide channel names in PubSub metrics
hide_stream_names=True, # Hide stream names in streaming metrics
# Custom histogram buckets
buckets_operation_duration=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_stream_processing_duration=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_connection_create_time=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_connection_wait_time=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
))
r = redis.Redis(host='localhost', port=6379)
r.set('key', 'value') # Metrics collected automatically
r.get('key')
otel.shutdown()
The available options for OTelConfig are described in the table below:
| Option | Type | Description |
|---|---|---|
metric_groups |
List[MetricGroup] |
List of metric groups to enable. By default, only CONNECTION_BASIC and RESILIENCY are enabled. See Redis metric groups for a list of available groups. |
include_commands |
List[str] |
List of Redis commands to track. If set, only these commands will be tracked. Note that you should use the Redis command name rather than the Python method name where the two differ. |
exclude_commands |
List[str] |
List of Redis commands to exclude from tracking. If set, all commands except these will be tracked. Note that you should use the Redis command name rather than the Python method name where the two differ. |
hide_pubsub_channel_names |
bool |
If true, channel names in pub/sub metrics will be hidden. |
hide_stream_names |
bool |
If true, stream names in streaming metrics will be hidden. |
buckets_operation_duration |
List[float] |
List of bucket boundaries for the operation.duration histogram (see Custom histogram buckets below). |
buckets_stream_processing_duration |
List[float] |
List of bucket boundaries for the stream.processing.duration histogram (see Custom histogram buckets below). |
buckets_connection_create_time |
List[float] |
List of bucket boundaries for the connection.create.time histogram (see Custom histogram buckets below). |
buckets_connection_wait_time |
List[float] |
List of bucket boundaries for the connection.wait.time histogram (see Custom histogram buckets below). |
Custom histogram buckets
For the histogram metrics, a reasonable default set of buckets is defined, but
you can customize the bucket boundaries to suit your needs (the buckets are the
ranges of data values counted for each bar of the histogram). Pass an increasing
list of float values to the buckets_xxx options when you create the OTelConfig
object. The first and last values in the list are the lower and upper bounds of the
histogram, respectively, and the values in between define the bucket boundaries.
Use Redis
Once you have configured the client, all Redis connections you create will be automatically instrumented and the collected metrics will be exported to your configured destination.
The example below shows the simplest Redis connection and a few commands, but see the observability demonstration repository for an example that calls a variety of commands in a more realistic way.
# OpenTelemetry metrics API
from opentelemetry import metrics
from opentelemetry.sdk.metrics import MeterProvider
from opentelemetry.sdk.metrics.export import PeriodicExportingMetricReader
from opentelemetry.exporter.otlp.proto.http.metric_exporter import OTLPMetricExporter
# Redis observability API
from redis.observability.providers import get_observability_instance
from redis.observability.config import OTelConfig, MetricGroup
# Redis client
import redis
exporter = OTLPMetricExporter(endpoint="http://localhost:4318/v1/metrics")
reader = PeriodicExportingMetricReader(exporter=exporter, export_interval_millis=10000)
provider = MeterProvider(metric_readers=[reader])
metrics.set_meter_provider(provider)
otel = get_observability_instance()
otel.init(OTelConfig(
# Metric groups to enable (default: CONNECTION_BASIC | RESILIENCY)
metric_groups=[
MetricGroup.CONNECTION_BASIC, # Connection creation time, relaxed timeout
MetricGroup.CONNECTION_ADVANCED, # Connection wait time, timeouts, closed connections
MetricGroup.COMMAND, # Command execution duration
MetricGroup.RESILIENCY, # Error counts, maintenance notifications
MetricGroup.PUBSUB, # PubSub message counts
MetricGroup.STREAMING, # Stream message lag
MetricGroup.CSC, # Client Side Caching metrics
],
# Filter which commands to track
include_commands=['GET', 'SET', 'HGET'], # Only track these commands
# OR
exclude_commands=['DEBUG', 'SLOWLOG'], # Track all except these
# Privacy controls
hide_pubsub_channel_names=True, # Hide channel names in PubSub metrics
hide_stream_names=True, # Hide stream names in streaming metrics
# Custom histogram buckets
buckets_operation_duration=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_stream_processing_duration=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_connection_create_time=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_connection_wait_time=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
))
r = redis.Redis(host='localhost', port=6379)
r.set('key', 'value') # Metrics collected automatically
r.get('key')
otel.shutdown()
Shutdown
When your application exits, you should call the shutdown() method to ensure
that all pending metrics are exported.
# OpenTelemetry metrics API
from opentelemetry import metrics
from opentelemetry.sdk.metrics import MeterProvider
from opentelemetry.sdk.metrics.export import PeriodicExportingMetricReader
from opentelemetry.exporter.otlp.proto.http.metric_exporter import OTLPMetricExporter
# Redis observability API
from redis.observability.providers import get_observability_instance
from redis.observability.config import OTelConfig, MetricGroup
# Redis client
import redis
exporter = OTLPMetricExporter(endpoint="http://localhost:4318/v1/metrics")
reader = PeriodicExportingMetricReader(exporter=exporter, export_interval_millis=10000)
provider = MeterProvider(metric_readers=[reader])
metrics.set_meter_provider(provider)
otel = get_observability_instance()
otel.init(OTelConfig(
# Metric groups to enable (default: CONNECTION_BASIC | RESILIENCY)
metric_groups=[
MetricGroup.CONNECTION_BASIC, # Connection creation time, relaxed timeout
MetricGroup.CONNECTION_ADVANCED, # Connection wait time, timeouts, closed connections
MetricGroup.COMMAND, # Command execution duration
MetricGroup.RESILIENCY, # Error counts, maintenance notifications
MetricGroup.PUBSUB, # PubSub message counts
MetricGroup.STREAMING, # Stream message lag
MetricGroup.CSC, # Client Side Caching metrics
],
# Filter which commands to track
include_commands=['GET', 'SET', 'HGET'], # Only track these commands
# OR
exclude_commands=['DEBUG', 'SLOWLOG'], # Track all except these
# Privacy controls
hide_pubsub_channel_names=True, # Hide channel names in PubSub metrics
hide_stream_names=True, # Hide stream names in streaming metrics
# Custom histogram buckets
buckets_operation_duration=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_stream_processing_duration=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_connection_create_time=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
buckets_connection_wait_time=[
0.0001, 0.00025, 0.0005, 0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1,
0.25, 0.5, 1, 2.5,
],
))
r = redis.Redis(host='localhost', port=6379)
r.set('key', 'value') # Metrics collected automatically
r.get('key')
otel.shutdown()