This plugin receives data from a Kafka broker and provides values for histogramming. In the Kafka context, this entails establishing consumers. However, from the perspective of daquiri, the plugin constitutes a producer. Once the plugin has interpreted a buffer according to the appropriate flatbuffer schema, the event data is delivered to the histogramming service via internal "streams".
Once you have added an "ESSStream" plugin to your profile configuration, you should set your Kafka broker:
You can now establish consumers for multiple Kafka topics. The "fast-forward" option allows you to start dropping packets if daquiri is not keeping up with the data rates. This may be desirable if you want to see the most up-to-date images, but you will have no guarantee that they are representative. If you wish to faithfully histogram the data, this option should be off.
For each topic you must select a flat buffer schema for parsing. There is currently no auto-detection of schema or multiplexing support implemented, so if you have different types of data incoming, you must stream them on separate topics.
Here is where you must provide an identifier string for the internal daquiri stream. These ids will reappear when configuring histograms to bin specific values. Depending on the chosen schema, a single Kafka topic may translate into multiple daquiri streams.
If your schema requires a logical geometry configuration, it must match that used by the EFU when producing the data. This tells the plugin how to translate pixel ids into coordinate values. The EFU's logical geometry configuration is most likely in the detector-specific json file.
When working with prototype detectors, timing information may not be fully available. If your does not provide valid offset (pulse) time for each data frame, you may want daquiri to keep a local "high clock" to ensure that timestamps are monotonically increasing.
If you are getting valid timing data and expect to have meaningful TOF (or other time-based) histograms, you should also configure the time-base of your clock. You will not be able to construct meaningful TOF histograms if pulse information is not provided by the EFU.
Here is an example of a schema that results in multiple daquiri data streams. Since these fields constitute independent sets of information, we separate them out for before binning. Make sure the stream identifiers are unique.
