CERN HSE Computing (HSE-TS-CS)

Contact email: hse-ts-cs@cern.ch

WinCC OA Kafka Driver

Table of contents

1. Description

2. Libraries

   2.1 Requirements for external libraries

3. Compilation

   3.1. To build the driver

   3.2. Build options

   3.3. To install and update

   3.4. Run

4. Config file

   4.1 Keytab (CERN Only)

5. WinCC OA Installation

6. Ready-to-launch project (CERN only)

   6.1. Project demo set-up

   • 6.1.1 Download & unzip
   • 6.1.2 Kafka Topic
   • 6.1.3 Keytab
   • 6.1.4 Config
   • 6.1.5 Install libraries
   • 6.1.6 Run the project

7. Kafka Driver Technical Documentation

   7.1. Main entry points

   7.2. Addressing DPEs with the Kafka driverl

   • 7.2.1 Data types
   • 7.2.2 For streaming
   • 7.2.3 For data ingestion
   • 7.2.4 Adding a new type

   7.3 Driver configuration

   7.4 Activity diagram

1. Description

This is a generic WinCC OA driver used to stream and/or ingest data via Kafka. The user has the following possibilities:

• Run a producer
• Run one or several consumers (one consumer per topic)
• Run both a producer and/or consumer(s)

2. Libraries

• C++11 STL
• WinCC OA API libraries
• cppkafka under BSD2 (check CMakeLists for release version)
• librdkafka under BSD2 (check CMakeLists for release version)

2.1 Requirements for external libraries

librdkafka and cppkafka are open source libraries, available on Github. They are both built from source via CMake.

You need to make sure to install the following libs on your system: * cyrus-sasl-gssapi * boost * cmake * openssl-dev

3. Compilation:

The project is CMake enabled. Note that if you set the PVSS_PROJ_PATH environment variable beforehand, you can install and update the driver via the install and update targets.

3.1 To build the driver

export PVSS_PROJ_PATH=<path_to_pvss_project_folder>
mkdir build
cd build
cmake ..
make -j

3.2 Build options

The default CMake build type is Release. The follow build types are taken into account:

Build Type	Details
Release	Default, 03 optimization
Debug	gdb enabled, no optimization
Coverage	gcov enabled, no optimization

---

To enable a specific build type, run:

cmake -DCMAKE_BUILD_TYPE=<build_type> ..

For Code Coverage make sure you have lcov installed. At the end of your tests, perform the following to get the html coverage report:

make coverage

If valgrind is installed, you can also run the tests with valgrind:

make valgrind

Once you are done with the tests, you can stop the valgrind process with:

make kill

3.3 To install and update

In order to install the driver in the PVSS_PROJ_PATH/bin, run:

make install

Making changes to the driver and updating it are simplified by using the update target, triggering a clean, install (and kill the driver):

make update

The kill target will simply terminate the running process. Combined with the always setting of the driver in WinCCOA, this will ensure that the driver is always up to date.

3.3 Run

It can be run from the WinCCOA Console or from command line and it will require the config.kafka driver_config_file:

./WINCCOAkafkaDrv -num <driver_number> -proj <project_name> +config <driver_config_file>

If you want to run the driver from the build folder (i.e. debug, coverage or valgrind mode) you can call:

make run

CMake will extract/parse the following information:

Variable	Details
project_name	The PVSS project name from PVSS_PROJ_PATH environment variable
driver_number	The driver number from the $PVSS_PROJ_PATH/config/progs file. Defaults to 999.
driver_config_file	The driver config file from the $PVSS_PROJ_PATH/config/progs file.

4. Config file

The config.kafka file has to be present under the WinCCOA project folder config.

Here, producer and consumer configs can be specified (based on your own needs). These will be parsed by the driver at startup thanks to the PRODUCER.CONFIG. keyword for producer and the CONSUMER.CONFIG. keyword for the consumer(s).

Here is an example config file:

[kafka]
PRODUCER.CONFIG.metadata.broker.list = dbnile-kafka-a-8.cern.ch:9093,dbnile-kafka-b-8.cern.ch:9093,dbnile-kafka-c-8.cern.ch:9093
PRODUCER.CONFIG.security.protocol = SASL_SSL
PRODUCER.CONFIG.sasl.mechanism = GSSAPI
PRODUCER.CONFIG.sasl.kerberos.service.name = kafka
PRODUCER.CONFIG.sasl.kerberos.principal = asavules
PRODUCER.CONFIG.sasl.kerberos.keytab = /home/asavules/Workspace/as.keytab
PRODUCER.CONFIG.group.id = test-consumer-group
PRODUCER.CONFIG.statistics.interval.ms = 60000
CONSUMER.CONFIG.metadata.broker.list = dbnile-kafka-a-8.cern.ch:9093,dbnile-kafka-b-8.cern.ch:9093,dbnile-kafka-c-8.cern.ch:9093
CONSUMER.CONFIG.security.protocol = SASL_SSL
CONSUMER.CONFIG.sasl.mechanism = GSSAPI
CONSUMER.CONFIG.sasl.kerberos.service.name = kafka
CONSUMER.CONFIG.sasl.kerberos.principal = asavules
CONSUMER.CONFIG.sasl.kerberos.keytab = /home/asavules/Workspace/as.keytab
CONSUMER.CONFIG.statistics.interval.ms = 60000
CONSUMER.CONFIG.group.id = remustest_meas_consumer
CONSUMER.CONFIG.enable.auto.commit = true
CONSUMER.CONFIG.auto.offset.reset = latest

In this config file you can add other producer or consumer configuration entries(see Kafka Official Documentation).

4.1 Keytab (CERN Only)

To be able to access kafka, you will need to create a keytab.

Please refer to official user guide, section How to generate a keytab file: CERN Kafka Official User Guide

For example with NICE user jdoe:

    cern-get-keytab --user --keytab jdoe.keytab --login jdoe

Edit the config/config.kafka file in your WinCC OA project and update the CONSUMER.CONFIG.sasl.kerberos.keytab and/or PRODUCER.CONFIG.sasl.kerberos.keytab entry to point to your keytab file + the CONSUMER.CONFIG.sasl.kerberos.principal and/or PRODUCER.CONFIG.sasl.kerberos.principal to your NICE username.

For example:

[kafka]
PRODUCER.CONFIG.metadata.broker.list = dbnile-kafka-a-8.cern.ch:9093,dbnile-kafka-b-8.cern.ch:9093,dbnile-kafka-c-8.cern.ch:9093
PRODUCER.CONFIG.security.protocol = SASL_SSL
PRODUCER.CONFIG.sasl.mechanism = GSSAPI
PRODUCER.CONFIG.sasl.kerberos.service.name = kafka
PRODUCER.CONFIG.sasl.kerberos.principal = jdoe
PRODUCER.CONFIG.sasl.kerberos.keytab = /home/jdoe/Workspace/jdoe.keytab
PRODUCER.CONFIG.group.id = test-producer-group-demo
PRODUCER.CONFIG.statistics.interval.ms = 60000

CONSUMER.CONFIG.metadata.broker.list = dbnile-kafka-a-8.cern.ch:9093,dbnile-kafka-b-8.cern.ch:9093,dbnile-kafka-c-8.cern.ch:9093
CONSUMER.CONFIG.security.protocol = SASL_SSL
CONSUMER.CONFIG.sasl.mechanism = GSSAPI
CONSUMER.CONFIG.sasl.kerberos.service.name = kafka
CONSUMER.CONFIG.sasl.kerberos.principal = jdoe
CONSUMER.CONFIG.sasl.kerberos.keytab = /home/jdoe/Workspace/jdoe.keytab
CONSUMER.CONFIG.statistics.interval.ms = 60000
CONSUMER.CONFIG.group.id = remustest_demo_consumer
CONSUMER.CONFIG.enable.auto.commit = true

5. WinCC OA Installation

Under the winccoa folder you will find the following files that you need to copy to your pre-existing project in the corresponding paths:

• dplist/kafka_driver_config.dpl : it contains CONFIG_KAFKA DPEs. It is used to set the driver options such as log levels and to retreive consumer and producer statistics. Once you've successfully launched the driver in the WinCC project management, you can import it via the ASCII Manager(refer to the official WinCC OA Documentation).

Notes:

* The internal driver number in the dump is 2. If it's unavailable to you, try to modify the dump file directly. 

• dplist/panels/para/address_kafka.pnl : a panel that you can use in para for kafka addressing. If you install this panel, then you will also need the WinCC OA scripts that go along:

  • scripts/userDrivers.ctl
  • scripts/userPara.ctl

In order to set up the addressing of kafka DPE form a control script, you may use the folowing library: * scripts/libs/kafka_dpe_addressing.ctl. Usage:

`kafkaAddress_addressDPE("Consumer.myDPE", kafkaAddress_DATA_TYPE_STRING, kafkaAddress_MODE_IN, 2, "my_kafka_topic", "my_kafka_key");`
`kafkaAddress_addressDPE("Producer.myDPE", kafkaAddress_DATA_TYPE_STRING, kafkaAddress_MODE_OUT, 2, "my_kafka_topic", "my_kafka_key", 100);`

See 7.3 Driver configuration section for a brief descprition of relevant CONFIG_KAFKA DPEs.

6. Ready-to-launch project (CERN only)

A ready-to-launch WinCCOA 3.16 kafka consumer demo project available here:

• winccoa316_demo_project/kafka_producer_consumer_demo.zip

6.1 Project demo set-up

6.1.1 Download & unzip

Download winccoa316_demo_project/kafka_producer_consumer_demo.zip.

Unzip it on a CentOS7 machine that have WinCCOA 3.16 installed. You will need sudo rights to install libraries.

6.1.2 Kafka Topic

The topic remustest_demo from where the values are retrieved in the demo is available on the CERN kafka-gptn cluster.

To be able to retrieve the values for the demo project, you need to get access for this topic. Please subscribe to the kafka-remustest-demo egroup.

6.1.3. Keytab

To be able to access kafka, you will need to create a keytab, to replace the empty one in config/ .

go to [PROJECT PATH]/config and run

cern-get-keytab --user --keytab user.keytab --login [NICE login]

See 4.1. Keytab (CERN Only) section for more details

6.1.4. Config

Edit the config/config.kafka file and update:

* `CONSUMER.CONFIG.sasl.kerberos.keytab` path to your keytab file (consumer).
* `CONSUMER.CONFIG.sasl.kerberos.principal` your CERN's NICE username (consumer).
* `PRODUCER.CONFIG.sasl.kerberos.keytab` path to your keytab file (producer).
* `PRODUCER.CONFIG.sasl.kerberos.principal` your CERN's NICE username (producer).

Edit the config/config file to point to the appropriate project path.

6.1.5. Install libraries

First run :

sudo yum install cyrus-sasl-gssapi

Then go to bin/libs/ and run:

./installLocal.sh

This will copy the libraries and update dependencies. Note that you need sudo rights.

6.1.6. Run the project

Simply register then start the project in WinCCOA.

A UI manager will launch para module, where producer and consumer DPEs are already set-up and running.

The manager -f kafka_producer_example.lst is continuously doing dpSets into the Producer DPEs, that are sending messages to the kafka topic "remustest_demo". The messages are transmitted to Kafka, then read by the Consumer DPEs.

Since the messages are going through Kafka, they can of course be read/write from external software (other WCCOA project, java, python, MQTT, ...) communicating with the same topic, as demonstrated on the screeshot, where messages are sent through the kafka command line tool (https://kafka.apache.org/quickstart)

(Optional) Running the manager -f kafka_addressing_example.lst will demonstrate a dynamic addressing of kafka DPE, both producers and consumers, from Control scripts.

7. Kafka Driver Technical Documentation

7.1 Main entry points

After the driver startup, the main entry points in the driver are:

• kafkaHwService::writeData() -> WinCC to Driver communication

  This is how the kafka streaming is performed. Thanks to the addressing <TOPIC>$<KEY>[$<DEBOUNCING_TIMEFRAME>], the driver will be able to stream to the right topic.

• kafkaHwService::workProc() -> Driver to WinCC communication

  This is how we push data to WinCC from Kafka. Thanks to the addressing <TOPIC>$<KEY>,the driver will be able to map the data ingested from the respective Kafka topic to the WinCC DPE.

Please refer to the WinCC documentation for more information on the WinCC OA API. For more info on the debouncing, see Remus RealTime Evolution - KAFKA presentation.

7.2 Addressing DPEs with the Kafka driver

7.2.1 Data Types

When the kafka driver pushes a DPE value to WinCC, a transformation takes place. See Transformations folder. We are currently supporting the following data types for the periphery address:

---

WinCC DataType	Transformation class	Periphery data type value
bool	kafkaBoolTrans.cxx	1000 (TransUserType def in WinCC OA API)
uint8	kafkaUint8Trans.cxx	1001 (TransUserType + 1)
int32	kafkaInt32Trans.cxx	1002 (TransUserType + 2)
int64	kafkaInt64Trans.cxx	1003 (TransUserType + 3)
float	kafkaFloatTrans.cxx	1004 (TransUserType + 4)
string	kafkaStringTrans.cxx	1005 (TransUserType + 5)
time	kafkaTimeTrans.cxx	1006 (TransUserType + 6)

---

7.2.2 For streaming

The periphery address data type has to be set to string (1005).

The direction in the periphery has to be set to OUT. The addressing is generic: <TOPIC>$<KEY>[$<DEBOUNCING_TIMEFRAME>]. The $DEBOUNCING_TIMEFRAME is optional and if not set then it is 0 by default.

• Example 1

  remustest_demo$MeasurementJSON123$200

  Gives us :

  • The kafka topic: remustest_demo
  • The key key: MeasurementJSON123
  • Debouncing timeframe: 200 milliseconds

• Example 2

  remustest_demo$MeasurementJSON123

  Gives us :

  • The kafka topic: remustest_demo
  • The key key: MeasurementJSON123
  • No debouncing timeframe

For more info on the addressing, see Remus RealTime Evolution - KAFKA presentation.

7.2.3 For data ingestion

The direction in the periphery address has to be set to IN. The addressing is generic: <TOPIC>$<KEY>.

For example

remustest_demo$MeasurementJSON123

Gives us :
* The kafka topic: remustest_demo
* The key key: MeasurementJSON123

7.2.4 Adding a new transformation

To add a new transformation you need to do the following:

• create a define in kafkaHWMappeer.hxx

    #define kafkaDrvDoubleTransType (TransUserType + 7)
  

• handle the new transformation type in kafkaHWMapper::addDpPa()

• implement the transformation type class. The important functions here are

  • ::toPeriph(...) for WinCC OA to Kafka driver transformation
  • ::toVar(...) for Kafka driver to WinCC OA transformation

7.3 Driver Configuration

Available via the WinCC OA CONFIG_KAFKA DataPoint, we have the following

Config DPE	Direction	Addressing	Type	Description
DebugLvl	OUT	DEBUGLVL	INT32	Debug Level for logging. You can use this to debug issues. (default 0)
DebouncingThreadInterval	OUT	DEBOUNCINGTHREADINTERVAL	INT32	In milliseconds. The sleep interval for the debouncing thread (default 50 ms)
MaxPollRecords	OUT	MAXPOLLRECORDS	INT32	MAximum number of records should the consumer poll retrieve (default 1000)
IN.ConsumerStatsDP	IN	CONSUMER_STATISTICS	STRING	This is where the driver periodically pushes statistics from kafka consumer (json)
IN.ProducerStatsDP	IN	PRODUCER_STATISTICS	STRING	This is where the driver periodically pushes statistics from kafka producer (json)
IN.ProducerAllBrokersDown	IN	PRODUCER_ALL_BROKERS_DOWN	BOOL	Set by the when all brokers are down for the producer.
IN.NoConsumerConfig	IN	CONSUMER_NO_CONFIG	BOOL	Indicates there is no consumer configuration available.
IN.NoProducerConfig	IN	PRODUCER_NO_CONFIG	BOOL	Indicates there is no producer configuration available.

7.4 Activity Diagram