Changes

Next

Remove support for deprecated ep00 schema
Adopt pip-tools to manage requirements.txt files
Refactor statistics reporter to support dynamically added metrics
Add latency and per-PV graphite metrics
Add containerfile

v2.1.0

Fix: Do not re-publish al00 alarms if state has not changed
Fix: In ep01 even if the PV was connected it appeared as unknown
Fix: Updating tdct to handle empty arrays coming from EPICS
Feat: Forward al00 messages for tdct, se00 and senv schemas
Feat: Export kafka_delivery_errors metric when Kafka producer permanently fails delivery
Feat: Add support for SASL_SSL in Kafka
Feat: Make the package installable.
Libs: Update confluent_kafka to v2.1

v2.0.0

Using ep01 instead of ep00 for EPICS connection events.
Adding the schema "se00", which is in an upgrade to the schema "senv" which could not handle floats/doubles.
Added support for f144 and al00 schemas.
Refactoring to separate Serialisers in classes that implement a typing.Protocol.

pre-2.0

Repetition of messages will now be triggered only if no new message has been received within the configurable repetition time (pv-update-period).
Drastic improvement to serialisation of f142 messages by updating to streaming-data-types 0.15.1
Fixed bug when serialising empty senv-messages.
Fixed bug where the wrong type was picked when casting numpy arrays in the f142 serialiser.
Known bad PV updates will no longer be cached for (attempted) re-transmission.
Some exceptions in the PVA/CA update handling code will now produce log messages.
Re-factored the code to allow automatic instantiation of forwarding modules
Added separate ep00 module for handling only connection status serialisation
New ep00 module is now automatically instantiated for every PV that the forwarder is configured to forward
System test running has been simplified and fixed
Set the "facility" and "process_name" fields in the Graylog messages to "ESS" and "forwarder"
System tests renamed to integration tests
Repeated messages will now have the Kafka timestamp set to the current system time. This in an effort to reduce the load on the Kafka broker.