Hi Chris, here a few thoughts:
*I haven't seen a single use case of someone writing to a PLC with PLC4X * => This sounds similar to most IoT MQTT scenarios I see today. The main challenge is to get data out of machines to analyze and use it. Writing back (e.g. to control the systems) is the second step which comes later, sometimes years later - or even never if not needed. * I doubt such a scenario is ideal for the Kafka setup* => If you don't see it in other PLC4X projects, then I would not worry about it too much for Kafka integration today. However, I would not think too much about Kafka here. Either you see many use cases for Sinks in general today for PLC4X, or you don't see them yet. Some use Kafka, some use Java, some use Nifi or whatever. *A PLC is not able to consume a lot of data until it chokes and stops working* => This is an argument we hear often in Kafka community. Often it is *not* valid (and we see more and more use cases where Kafka is just used for transactional data like bank payments which are very low volume). Therefore, Kafka is not just used for big data sets. In IoT scenarios (like connected cars), we see more and more use cases where people also want to send information back to the device (e.g. alerts or recommendations to the driver, though, this is just a few messages, not comparable to sensor egress messages). In IoT scenarios, in almost all cases, you ingest a lot of IoT data from the devices / machines to other clusters (kafka, spark, cloud, AI, whatever), but if you send data back to the device / machine, it is just control data or other limited, small data sets. So a high throughput egress from IoT devices does not imply a high throughput ingress. Therefore, I recommend to not use this argument. * So in a first round I'll concentrate on implementing a robust Scraper-based PLC4X Kafka Connect Source, but drop the Sink entirely. If we find out that a Sink makes sense we'll be able to add it in a later version* I would recommend exactly the same. Focus on the important direction, and implement the other direction later (if people ask for it and share their use cases). Confluent did exactly the same for many other connectors, e.g. AWS S3 Sink or MQTT Source. Then we also built MQTT Sink and S3 Source later because we saw more demand for these, too. Best regards, Kai Waehner On Tue, Aug 13, 2019 at 3:52 PM Christofer Dutz <christofer.d...@c-ware.de> wrote: > Hi all, > > so while I was working on the Kafka Sink I sort of started thinking if > such a thing is a good idea to have at all. > A Kafka system would be able to provide a vast stream of data that would > have to be routed to a PLCs. > On the one side, I haven't seen a single usecase of someone writing to a > PLC with PLC4X and I doubt such a scenario is ideal for the Kafka setup. > A PLC is not able to consume a lot of data until it chokes and stops > working. > > So in a first round I'll concentrate on implementing a robust > Scraper-based PLC4X Kafka Connect Source, but drop the Sink entirely. > > If we find out that a Sink makes sense we'll be able to add it in a later > version. > > What do you think? > > Chris > > > Am 02.08.19, 11:24 schrieb "Julian Feinauer" < > j.feina...@pragmaticminds.de>: > > Hi, > > I think that's the best way then. > > I agree with your suggestion. > > Julian > > Von meinem Mobiltelefon gesendet > > > -------- Ursprüngliche Nachricht -------- > Betreff: Re: [KAFKA] Refactoring the Kafka Connect plugin? > Von: Christofer Dutz > An: dev@plc4x.apache.org > Cc: > > Hi all, > > Ok so it seems that Kafka only supports configuration via simple > unstructured maps. > So if we want some hierarchical configuration like the proposed one, > we'll have to do it log4j.properties-style: > > name=plc-source-test > connector.class=org.apache.plc4x.kafka.Plc4xSourceConnector > > defaults.default-topic=some/default > > sources.machineA.url=s7://1.2.3.4/1/1 > sources.machineA.jobs.s7-dashboard.enabled=true > sources.machineA.jobs.s7-heartbeat.enabled=true > sources.machineA.jobs.s7-heartbeat.topic=heartbeat > > sources.machineB.url=s7://10.20.30.40/1/1 > sources.machineB.topic=heartbeat > sources.machineB.jobs.s7-heartbeat.enabled=true > > sources.machineC.url=ads://1.2.3.4.5.6 > sources.machineC.topic=heartbeat > sources.machineC.jobs.ads-heartbeat.enabled=true > > jobs.s7-dashboard.scrapeRate=500 > jobs.s7-dashboard.fields.inputPreasure=%DB.DB1.4:INT > jobs.s7-dashboard.fields.outputPreasure=%Q1:BYT > jobs.s7-dashboard.fields.temperature=%I3:INT > > jobs.s7-heartbeat.scrapeRate=1000 > jobs.s7-heartbeat.fields.active=%I0.2:BOOL > > jobs.ads-heartbeat.scrapeRate=1000 > jobs.ads-heartbeat.active=Main.running > > > Chris > > > > Am 01.08.19, 15:46 schrieb "Christofer Dutz" < > christofer.d...@c-ware.de>: > > Hi Kai, > > that document is exactly the one I'm currently using. > What I'm currently working on is updating the current plugin to > not schedule and handle the connection stuff manually, but use the scraper > component of PLC4X. > Also is the current configuration not production ready and I'll be > working on to make it more easily usable. > > But it will definitely not hurt to have some Kafka Pro have a look > at what we did and propose improvements. After all we want the thing to be > rock-solid :-) > > Chris > > > > Am 31.07.19, 17:03 schrieb "Kai Wähner" <megachu...@gmail.com>: > > Hi Chris, > > great that you will work on the connector. > > I am not deep technical, but if you need guidance from Kafka > Connect > experts, I can connect you to a Confluent colleague to can > help with best > practices for building the connector. > > For example, we have implemented a wildcard option into our > MQTT Connector > to map MQTT Topics to Kafka Topics in a more flexible way > (e.g. 1000s of > cars from different MQTT Topics can be routed into 1 Kafka > Topic). This > might also be interesting for this connector as you expect to > various PLCs. > > This guide might also help: > > https://www.confluent.io/wp-content/uploads/Verification-Guide-Confluent-Platform-Connectors-Integrations.pdf > > > > On Wed, Jul 31, 2019 at 4:39 PM Christofer Dutz < > christofer.d...@c-ware.de> > wrote: > > > Hi all, > > > > I am currently planning on cleaning up the Kafka Connect > adapter a little > > as this was implemented as part of a proof of concept and is > still I a > > state I wouldn’t use in production ;-) > > But a lot has happened since then and I’m planning on making > it a really > > usable tool in the next few days. > > > > A lot has changed since we created the integration module > QT3 2018 and I > > would like to refactor it to use the Scraper for the heavy > lifting. > > > > Currently a user has to provide a parameter “query” which > contains a > > comma-separated list of connection-strings with appended > address. This is > > purely unmanageable. > > > > I would like to make it configurable via JSON or Yaml file. > > > > I think it would make sense to define groups of fields that > are collected > > on one device at an equal rate. So it’s pretty similar to > the scraper > > example, however I would like to not specify the source in > the job, but the > > other way around. > > When specifying the “sources” I would also provide which > jobs should run > > on a given collection. > > As the connector was initially showcased in a scenario where > data had to > > be collected on a big number of PLCs with equal specs, > > I think this is the probably most important use-case and in > this it is > > also probably more common to add new devices to collect > standard data on > > than the other way around. > > > > Also should we provide the means to also set per connection > to which > > kafka-topic the data should be sent to. > > We could provide the means to set a default and make it > optional however. > > When posting to a topic we also need to provide means for > partitioning, so > > I would provide sources with an optional “name”. > > Each message would not only have the data requested, but > also the > > source-url, source-name and the job-name with a timestamp. > > > > So I guess it would look something like this: > > > > # > > > ---------------------------------------------------------------------------- > > # Licensed to the Apache Software Foundation (ASF) under one > > # or more contributor license agreements. See the NOTICE > file > > # distributed with this work for additional information > > # regarding copyright ownership. The ASF licenses this file > > # to you under the Apache License, Version 2.0 (the > > # "License"); you may not use this file except in compliance > > # with the License. You may obtain a copy of the License at > > # > > # http://www.apache.org/licenses/LICENSE-2.0 > > # > > # Unless required by applicable law or agreed to in writing, > > # software distributed under the License is distributed on an > > # "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY > > # KIND, either express or implied. See the License for the > > # specific language governing permissions and limitations > > # under the License. > > # > > > ---------------------------------------------------------------------------- > > --- > > # Defaults used throughout all collections > > defaults: > > # If not specified, all data goes to this topic (optional) > > default-topic: some/default > > > > # Defines connections to PLCs > > sources: > > # Connection to a S7 device > > - name: machineA > > # PLC4X connection URL > > url: s7://1.2.3.4/1/1 > > jobs: > > # Just references the job "s7-dashboard". All data > will be published > > to the default topic > > - name: s7-dashboard > > # References the job "s7-heartbeat", however is > configures the > > output to go to the topic "heartbeat" > > - name: s7-heartbeat > > topic: heartbeat > > > > # Connection to a second S7 device > > - name: machineB > > url: s7://10.20.30.40/1/1 > > # Sets the default topic for this connection. All jobs > data will go to > > "heartbeat" > > topic: heartbeat > > jobs: > > - s7-heartbeat > > > > # Connection to a Beckhoff device > > - name: machineC > > url: ads://1.2.3.4.5.6 > > topic: heartbeat > > jobs: > > - ads-heartbeat > > > > # Defines what should be collected how often > > jobs: > > # Defines a job to collect a set of fields on s7 devices > every 500ms > > - name: s7-dashboard > > scrapeRate: 500 > > fields: > > # The key will be used in the Kafka message to > identify this field, > > the value here contains the PLC4X address > > inputPreasure: %DB.DB1.4:INT > > outputPreasure: %Q1:BYTE > > temperature: %I3:INT > > > > # Defines a second job to collect a set of fields on s7 > devices every > > 1000ms > > - name: s7-heartbeat > > scrapeRate: 1000 > > fields: > > active: %I0.2:BOOL > > > > # Defines a third job that collects data on Beckhoff > devices > > - name: ads-heartbeat > > scrapeRate: 1000 > > fields: > > active: Main.running > > > > I think it should be self-explanatory with my comments > inline. > > > > What do you think? > > > > Chris > > > > > > > > > > >
