Flink streaming job is not scaling as expected - apache-flink

We are in the middle of testing scaling ability of Flink. But we found that scaling not working, no matter increase more slot or increase number of Task Manager. We would expect a linear, if not close-to-linear scaling performance but the result even show degradation. Appreciated any comments.
Test Details,
-VMWare vsphere
-Just a simple pass through test,
- auto gen source 3mil records, each 1kb in size, parallelism=1
- source pass into next map operator, which just return the same record, and sent counter to statsD, parallelism is in cases = 2,4,6
3 TM, total 6 slots(2/TM) each JM/TM has 32 vCPU, 100GB memory
Result:
2 slots: 26 seconds, 3mil/26=115k TPS
4 slots: 23 seconds, 3mil/23=130k TPS
6 slots: 22 seconds, 3mil/22=136k TPS
As shown the scaling is almost nothing. Any clue? Thanks.

You really should be using a RichParallelSourceFunction. If you care about making the records from different instances of the source distinct, you can get ahold of each instance's index from the RuntimeContext, which is available via the getRuntimeContext() method in the RichFunction interface.
Also, Flink has a built-in statsd metrics reporter that you should be using instead of rolling your own. Moreover, numRecordsIn, numRecordsOut, numRecordsInPerSecond, and numRecordsOutPerSecond are already being computed for you, so no need to create this instrumentation yourself. You can also access these metrics via Flink's web interface, or the REST API.
As for why you might be experiencing poor scalability with the Kafka consumer, there are many things that could cause this. If you are using event time processing, then idle partitions could be holding things up (see https://issues.apache.org/jira/browse/FLINK-5479). If the stream is keyed, then data skew could be an issue. If you are connecting to an external database or service, then it could easily be a bottleneck. If checkpointing is misconfigured it could cause this. Or insufficient network capacity.
I would start to debug this by looking at some key metrics in the Flink web UI. Is the load well balanced across the sub-tasks, or is it skewed? You could turn on latency tracking and see if one of the kafka partitions is misbehaving (by inspecting the latency at the sink(s), which will be reported on a per-partition basis). And you could look for back pressure.

please refer to the sample code,
public class passthru extends RichMapFunction<String, String> {
public void open(Configuration configuration) throws Exception {
... ...
stats = new NonBlockingStatsDClient();
}
public String map(String value) throws Exception {
... ...
stats.increment();
return value;
}
}
public class datagen extends RichSourceFunction<String> {
... ...
public void run(SourceContext<String> ctx) throws Exception {
int i = 0;
while (run){
String idx = String.format("%09d", i);
ctx.collect("{\"<a 1kb json content with idx in certain json field>\"}");
i++;
if(i == loop)
run = false;
}
}
... ...
}
public class Job {
public static void main(String[] args) throws Exception {
... ...
DataStream<String> stream = env.addSource(new datagen(loop)).rebalance();
DataStream<String> convert = stream.map(new passthru(statsdUrl));
env.execute("Flink");
}
}
the reductionState code,
dataStream.flatMap(xxx).keyBy(new KeySelector<xxx, AggregationKey>() {
public AggregationKey getKey(rec r) throws Exception {
... ...
}
}).process(new Aggr());
public class Aggr extends ProcessFunction<rec, rec> {
private ReducingState<rec> store;
public void open(Configuration parameters) throws Exception {
store= getRuntimeContext().getReducingState(new ReducingStateDescriptor<>(
"reduction store", new ReduceFunction<rec>() {
... ...
}
public void processElement(rec r, Context ctx, Collector<rec> out)
throws Exception {
... ...
store.add(r);

Related

why data is not processed in RichFlatMapFunction

In order to improve the performance of data process, we store events to a map and do not process them untill event count reaches 100.
in the meantime, start a timer in open method, so data is processed every 60 seconds
this works when flink version is 1.11.3,
after upgrading flink version to 1.13.0
I found sometimes events were consumed from Kafka continuously, but were not processed in RichFlatMapFunction, it means data was missing.
after restarting service, it works well, but several hours later the same thing happened again.
any known issue for this flink version? any suggestions are appreciated.
public class MyJob {
public static void main(String[] args) throws Exception {
...
DataStream<String> rawEventSource = env.addSource(flinkKafkaConsumer);
...
}
public class MyMapFunction extends RichFlatMapFunction<String, String> implements Serializable {
#Override
public void open(Configuration parameters) {
...
long periodTimeout = 60;
pool.scheduleAtFixedRate(() -> {
// processing data
}, periodTimeout, periodTimeout, TimeUnit.SECONDS);
}
#Override
public void flatMap(String message, Collector<String> out) {
// store event to map
// count event,
// when count = 100, start data processing
}
}
You should avoid doing things with user threads and timers in Flink functions. The supported mechanism for this is to use a KeyedProcessFunction with processing time timers.

how to flush batch data to sink in apache flink

I am using apache flink(v1.10.0) to compute RabbitMQ message, the sink the result to MySQL, now I am compute like this:
consumeRecord.keyBy("gameType")
.timeWindowAll(Time.seconds(5))
.reduce((d1, d2) -> {
d1.setRealPumpAmount(d1.getRealPumpAmount() + d2.getRealPumpAmount());
d1.setPumpAmount(d1.getPumpAmount() + d2.getPumpAmount());
return d1;
})
.addSink(new SinkFunction<ReportPump>() {
#Override
public void invoke(ReportPump value, Context context) throws Exception {
// save to mysql
}
});
But now the sink method only get one row in each invoke, if one of rows in this batch failed,I could not rollback the batch operate.Now I want to get batch of one window and sink to database once, if failed, I rollback the insert and Apache Flink's checkpoint.This is what I trying to do now:
consumeRecord.keyBy("gameType")
.timeWindowAll(Time.seconds(5)).reduce(new ReduceFunction<ReportPump>() {
#Override
public ReportPump reduce(ReportPump d1, ReportPump d2) throws Exception {
d1.setRealPumpAmount(d1.getRealPumpAmount() + d2.getRealPumpAmount());
d1.setPumpAmount(d1.getPumpAmount() + d2.getPumpAmount());
return d1;
}
})
.apply(new AllWindowFunction<ReportPump, List<ReportPump>, TimeWindow>() {
#Override
public void apply(TimeWindow window, Iterable<ReportPump> values, Collector<List<ReportPump>> out) throws Exception {
ArrayList<ReportPump> employees = Lists.newArrayList(values);
if (employees.size() > 0) {
out.collect(employees);
}
}
})
.addSink(new SinkFunction<List<ReportPump>>() {
#Override
public void invoke(List<ReportPump> value, Context context) throws Exception {
PumpRealtimeHandler.invoke(value);
}
});
but the apply function give tips: Cannot resolve method 'apply' in 'SingleOutputStreamOperator'. How to reduce it and get list of the batch data and flush to database only once?
SingleOutputStreamOperator does not have an apply method, because apply can be emit only after windowing.
What you miss here is:
.windowAll(GlobalWindows.create())
between the reduce and the apply, it'll aggregate all the reduced results to one global window that contain list of all the reduces results, than you'll can make collect for one list instead multiple batchs against the database.
I dont know if your result is a good practice, because you'll lose the parallelism of apache flink.
You should read about the TableApi and the JDBC sink, maybe it'll help you. (more information about it here: https://ci.apache.org/projects/flink/flink-docs-stable/dev/table/connect.html#jdbc-connector).

Apache Flink - Counter value displayed but meter values not displayed

We are using Flink 1.8.0 and running it on EMR - Yarn and would like to measure the throughput.
Because our operators are chained, we have added meters and counters in our code - essentially an async operator that makes API calls with kinesis as both source and sync. In the Application Master i.e. Flink's web UI, we are able to get the value for the counters but not the meters.
public class AsyncClass extends RichAsyncFunction<String, String> {
private transient Counter counter;
private transient Meter meter;
#Override
public void open(Configuration parameters) throws Exception {
super.open(parameters);
this.counter = getRuntimeContext()
.getMetricGroup()
.counter("myCounter");
this.meter = getRuntimeContext()
.getMetricGroup()
.meter("myMeter", new DropwizardMeterWrapper(new com.codahale.metrics.Meter()));
}
#Override
public void close() throws Exception {
super.close();
ExecutorUtils.gracefulShutdown(20000, TimeUnit.MILLISECONDS, executorService);
}
#Override
public void asyncInvoke(String key, final ResultFuture<String> resultFuture) throws Exception {
resultFuture.complete(key);
this.meter.markEvent();
this.counter.inc();
}
}
To measure the complete throughput of the application, we obviously need the throughput of all the task managers together. Using meters, we are able to get the metrics for individual task managers. Is there any way to measure it at the operator level?
Turns out the meter displays whole number values and the rate is measured in decimals. When my load was a constant 1 event per second, it was actually measured as 0.9xxx something and hence was showing only 0 events per second.

use timeWindow and processingTime but data Ingestion was blockd

When I use Windows All, because there is only one degree of parallelism, there is a bottleneck in processing. Therefore, I change to timeWindow and use processTime, but I encounter a new problem, data can not be ingested. From the log on the console, it can be seen that only more than ten data are processed every second, if I use Windows All. It can process tens of thousands of data per second. So I don't know why.
When I added waterMark to time Windows, I found that time Windows can handle a large number of data per second, but upstream data still accumulates
SingleOutputStreamOperator<DataSetPOJO> dataSetPOJOSingleOutputStreamOperator = sdkInfos.flatMap(...);
dataSetPOJOSingleOutputStreamOperator.keyBy(new KeySelector<DataSetPOJO, String>() {
#Override
public String getKey(DataSetPOJO dataSet) {
return dataSet.getPartitionKey();
}
}).timeWindow(Time.seconds(3))
.process(new ProcessWindowFunction<DataSetPOJO, List<DataSetPOJO>, String, TimeWindow>() {
#Override
public void process(String key, Context context, Iterable<DataSetPOJO> elements,
Collector<List<DataSetPOJO>> out) throws Exception {
ArrayList<DataSetPOJO> dataSetPOJO = Lists.newArrayList(elements);
if (dataSetPOJO.size() > 0) {
// log.info("key~~~~~~~~~~~~~~:" + key);
// log.info("dataSetPOJO.size():" + dataSetPOJO.size());
out.collect(dataSetPOJO);
}
}
}).addSink(new Sink2Postgre());
I hope I can save enough batches in windows to write PostgreSQL,If this is not correct, how to write, if it is no problem, what will be the problem. Fink Version 1.5.3

getting numOfRecordsIn using counters in Flink

I want to show numRecordsIn for an operator in Flink and for doing this I have been following ppt by data artisans at here. code for the counter is given below
public static class mapper extends RichMapFunction<String,String>{
public Counter counter;
#Override
public void open(Configuration parameters) throws Exception {
super.open(parameters);
this.counter = getRuntimeContext()
.getMetricGroup()
.counter("numRecordsIn");
}
#Override
public String map(String s) throws Exception {
counter.inc();
System.out.println("counter val " + counter.toString());
return null;
}
}
The problem is that how do I specify which operator I want to show number_of_Records_In?
Metric counter are exposed via Flink's metric system. In order to take a look at them, you have to configure a metric reporter. A description how to register a metric reporter can be found here.
Flink includes a number of built-in metrics, including numRecordsIn. So if that's what you want to measure, there's no need to write any code to implement that particular measurement. Similarly for numRecordsInPerSecond, and a host of others.
The code you asked about causes the numRecordsIn counter to be incremented for the operator in which the metric is being used.
A good way to better understand the metrics system is to bring up a simple streaming job and look at the metrics in Flink's web ui. I also found it really helpful to query the monitoring REST api while a job was running.

Resources