Forwarder

Forwards EPICS PVs to Apache Kafka. Part of the ESS data streaming pipeline.

Both Channel Access and PV Access EPICS protocols are supported. The protocol can be configured independently for each PV.

Installing dependencies

Python 3.8 or higher is required. https://www.python.org/downloads/

Runtime Python dependencies are declared in pyproject.toml and pinned in requirements.txt at the root of the repository. They can be installed from a terminal by running

pip install -r requirements.txt

Launch

To run with minimal settings:

forwarder_launch.py --config-topic some_server:9092/some_config_topic --status-topic some_server:9092/some_status_topic --output-broker some_other_server:9092

For help:

forwarder_launch.py --help

Required arguments:

• config-topic - Kafka username/broker/topic (sasl_mechanism\user@broker:port/topic) to listen for commands relating to PVs to be forwarded on
• status-topic - Kafka username/broker/topic (sasl_mechanism\user@broker:port/topic) to publish regular status updates on
• output-broker - Kafka username/broker (sasl_mechanism\user@broker:port) to forward PV data into

Optional arguments:

• config-topic-sasl-password - Password for SASL Kafka authentication. Note that the username is specified in the config-topic argument
• status-topic-sasl-password - Password for SASL Kafka authentication. Note that the username is specified in the status-topic argument
• output-broker-sasl-password - Password for SASL Kafka authentication. Note that the username is specified in the output-broker argument
• storage-topic - Kafka username/broker/topic for storage of the current forwarding details; these will be reapplied when the forwarder is restarted
• storage-topic-sasl-password - Password for SASL Kafka authentication. Note that the username is specified in the storage-topic argument
• skip-retrieval - do not reapply stored forwarding details on start-up
• graylog-logger-address - Graylog logger instance to log to
• log-file - name of the file to log to
• pv-update-period - period for forward PVs values even if the value hasn't changed (milliseconds)
• service-id - identifier for this particular instance of the Forwarder
• fake-pv-period - period for random generated PV updates when channel_provider_type is set to 'fake' (milliseconds)

Arguments can also be specified in a configuration file

forwarder_launch --config-file my_config.conf

The configuration file consists of 'key=value' pairs, e.g.

output-broker=localhost:9092
pv-update-period=1000

Kafka authentication

Authentication is disabled by default. To enable it, you must provide the SASL mechanism, username and password as arguments.

The supported Kafka authentication SASL mechanisms are SCRAM-SHA-256, SCRAM-SHA-512 and PLAIN. Only non-TLS channels (SASL_PLAINTEXT) are currently supported.

The SASL mechanism can be specified as part of the username/broker string as follows: sasl_mechanism\username@broker:port/topic. Example: SCRAM-SHA-256\[email protected]:9092/topic.

Configuring EPICS PVs to be forwarded

Adding or removing PVs to be forwarded is done by publishing configuration change messages to the configuration topic specified in the command line arguments. Such messages must be serialised as FlatBuffers using the rf5k schema which can be found here. Support for serialising and deserialising these messages in Python in available in the ess-streaming-data-types library.

The schema takes two parameters: the type of configuration change (UpdateType) and the corresponding streams.

There are three choices for the UpdateType of the configuration message:

• ADD - add the specified streams to the existing set of streams
• REMOVE - remove the specified streams from the set of streams
• REMOVEALL - remove all streams

A stream contains:

• The name of the PV to be forwarded.
• The EPICS protocol for reading the PV.
• The Kafka topic to write to.
• The FlatBuffers schema to encode the PV value with. See the supported schemas here.
  • Note that additional messages with different schemas may be configured by the forwarder automatically. In particular, every PV will also generate ep01 messages, and PVs configured for f144 will forward alarm information as al00 messages.

Note that when removing (using REMOVE) configured streams, not all fields in the Stream table of the schema need to be populated. Missing or empty strings in the channel name, output topic and schema fields match all stream configurations. At least one field must be populated though. Single-character "?"" and multi-character "*" wildcards are allowed to be used in the channel name and output topic fields. In conjunction with naming conventions for EPICS channel names and Kafka topics this can be used to carry out operations such as clearing all configured streams for a particular instrument.

Empty PV updates are not forwarded and are not cached to send in periodic updates. This addresses, for example, the case of empty chopper timestamp updates when a chopper is not spinning.

Chopper timestamps to be forwarded with tdct schema are assumed to be in nanoseconds and relative to the EPICS update timestamp, they are converted to nanosecond-precision unix timestamps when forwarded.

A Python example

To use for real, replace CONFIG_BROKER, CONFIG_TOPIC and STREAMS with values corresponding to the real system.

from confluent_kafka import Producer
from streaming_data_types.forwarder_config_update_rf5k import (
    serialise_rf5k,
    StreamInfo,
    Protocol,
)
from streaming_data_types.fbschemas.forwarder_config_update_rf5k.UpdateType import (
    UpdateType,
)

CONFIG_BROKER = "some_kafka_broker:9092"

CONFIG_TOPIC = "TEST_forwarderConfig"

STREAMS = [
    StreamInfo("IOC:PV1", "f142", "some_topic", Protocol.Protocol.PVA),
    StreamInfo("IOC:PV2", "f142", "some_other_topic", Protocol.Protocol.CA),
    StreamInfo("IOC:PV3", "f142", "some_other_topic", Protocol.Protocol.PVA),
]

producer = Producer({"bootstrap.servers": CONFIG_BROKER})

# Add new streams
producer.produce(CONFIG_TOPIC, serialise_rf5k(UpdateType.ADD, STREAMS))

# Remove one stream (note that you need to pass the argument as a list)
# producer.produce(CONFIG_TOPIC, serialise_rf5k(UpdateType.REMOVE, [STREAMS[0]]))

# Remove all the streams at once
#   USE WITH CAUTION!!
# producer.produce(CONFIG_TOPIC, serialise_rf5k(UpdateType.REMOVEALL, []))

producer.flush()

Developer information

pre-commit hooks

black, flake8 and mypy can be used as a pre-commit hook (installed by pre-commit). You need to run

pre-commit install

once to activate the pre-commit check. To test the hooks run

pre-commit run --all-files

This command can also be used to run the hooks manually.

Installing development dependencies

Development dependencies (including all runtime dependencies) can be installed by using the following command

pip install -r requirements-dev.txt

Updating dependencies

The requirements.txt and requirements-dev.txt files are generated using pip-tools, which produces a full list of pinned versions generated from the dependencies declared in pyproject.toml.

To generate or update the complete list of dependencies:

1. Activate your local virtual environment and make sure you have pip and pip-tools installed: pip install -U pip pip-tools
2. Generate requirements.txt: pip-compile -v --resolver=backtracking -o requirements.txt pyproject.toml
3. Generate requirements-dev.txt: pip-compile -v --resolver=backtracking --extra dev -o requirements-dev.txt pyproject.toml

Developer notes

Additional design and implementation details are available at Developer notes.