I have a flow that executes spark jobs on Dataproc clusters in parallel for different zones. For each zone it creates a cluster, execute the spark job and delete the cluster after it finishes.
The spark job uses the org.apache.spark.rdd.PairRDDFunctions.saveAsNewAPIHadoopDataset method passing the BigQuery Configuration to save data on BigQuery table. The job saves data in more than one table, calling the saveAsNewAPIHadoopDataset method more than one time per job.
The problem is that sometimes I'm getting a error caused by a conflict in the Hadoop temporary BigQuery Dataset that it internally creates to run the jobs:
Exception in thread "main" com.google.api.client.googleapis.json.GoogleJsonResponseException: 409 Conflict
{
"code" : 409,
"errors" : [ {
"domain" : "global",
"message" : "Already Exists: Dataset <my-gcp-project>:<MY-DATASET>_hadoop_temporary_job_201802250620_0013",
"reason" : "duplicate"
} ],
"message" : "Already Exists: Dataset <my-gcp-project>:<MY-DATASET>_hadoop_temporary_job_201802250620_0013"
}
at com.google.api.client.googleapis.json.GoogleJsonResponseException.from(GoogleJsonResponseException.java:145)
at com.google.api.client.googleapis.services.json.AbstractGoogleJsonClientRequest.newExceptionOnError(AbstractGoogleJsonClientRequest.java:113)
at com.google.api.client.googleapis.services.json.AbstractGoogleJsonClientRequest.newExceptionOnError(AbstractGoogleJsonClientRequest.java:40)
at com.google.api.client.googleapis.services.AbstractGoogleClientRequest$1.interceptResponse(AbstractGoogleClientRequest.java:321)
at com.google.api.client.http.HttpRequest.execute(HttpRequest.java:1056)
at com.google.api.client.googleapis.services.AbstractGoogleClientRequest.executeUnparsed(AbstractGoogleClientRequest.java:419)
at com.google.api.client.googleapis.services.AbstractGoogleClientRequest.executeUnparsed(AbstractGoogleClientRequest.java:352)
at com.google.api.client.googleapis.services.AbstractGoogleClientRequest.execute(AbstractGoogleClientRequest.java:469)
at com.google.cloud.hadoop.io.bigquery.BigQueryOutputCommitter.setupJob(BigQueryOutputCommitter.java:107)
at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsNewAPIHadoopDataset$1.apply$mcV$sp(PairRDDFunctions.scala:1150)
at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsNewAPIHadoopDataset$1.apply(PairRDDFunctions.scala:1078)
at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsNewAPIHadoopDataset$1.apply(PairRDDFunctions.scala:1078)
at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151)
at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:112)
at org.apache.spark.rdd.RDD.withScope(RDD.scala:358)
at org.apache.spark.rdd.PairRDDFunctions.saveAsNewAPIHadoopDataset(PairRDDFunctions.scala:1078)
at org.apache.spark.api.java.JavaPairRDD.saveAsNewAPIHadoopDataset(JavaPairRDD.scala:819)
...
The timestamp 201802250620_0013 on the exception above has the _0013 sufix that I'm unsure if it represents time.
My thoughts is that sometimes the jobs runs at the same time and try to create a dataset with same timestamp in the name. Either in a parallel job or inside the same job on another saveAsNewAPIHadoopDataset call.
How can we avoid this error without putting a delay on the job execution?
The dependency that I'm using is:
<dependency>
<groupId>com.google.cloud.bigdataoss</groupId>
<artifactId>bigquery-connector</artifactId>
<version>0.10.2-hadoop2</version>
<scope>provided</scope>
</dependency>
The Dataproc image version is 1.1
Edit 1:
I tried using IndirectBigQueryOutputFormat but now I'm getting an error saying that the gcs output path already exists even passing it different time on each saveAsNewAPIHadoopDataset call.
Here is my code:
SparkConf sc = new SparkConf().setAppName("MyApp");
try (JavaSparkContext jsc = new JavaSparkContext(sc)) {
JavaPairRDD<String, String> filesJson = jsc.wholeTextFiles(jsonFolder, parts);
JavaPairRDD<String, String> jsons = filesJson.flatMapToPair(new FileSplitter()).repartition(parts);
JavaPairRDD<Object, JsonObject> objsJson = jsons.flatMapToPair(new JsonParser()).filter(t -> t._2() != null).cache();
objsJson
.filter(new FilterType(MSG_TYPE1))
.saveAsNewAPIHadoopDataset(createConf("my-project:MY_DATASET.MY_TABLE1", "gs://my-bucket/tmp1"));
objsJson
.filter(new FilterType(MSG_TYPE2))
.saveAsNewAPIHadoopDataset(createConf("my-project:MY_DATASET.MY_TABLE2", "gs://my-bucket/tmp2"));
objsJson
.filter(new FilterType(MSG_TYPE3))
.saveAsNewAPIHadoopDataset(createConf("my-project:MY_DATASET.MY_TABLE3", "gs://my-bucket/tmp3"));
// here goes another ingestion process. same code as above but diferrent params, parsers, etc.
}
Configuration createConf(String table, String outGCS) {
Configuration conf = new Configuration();
BigQueryOutputConfiguration.configure(conf, table, null, outGCS, BigQueryFileFormat.NEWLINE_DELIMITED_JSON, TextOutputFormat.class);
conf.set("mapreduce.job.outputformat.class", IndirectBigQueryOutputFormat.class.getName());
return conf;
}
I believe what may be happening is that each mapper tries to create its own dataset. This is rather inefficient (and burns your daily quota proportional to the number of mappers).
An alternative is to use IndirectBigQueryOutputFormat for output class:
IndirectBigQueryOutputFormat works by first buffering all the data into a Cloud Storage temporary table, and then, on commitJob, copies all data from Cloud Storage into BigQuery in one operation. Its use is recommended for large jobs since it only requires one BigQuery "load" job per Hadoop/Spark job, as compared to BigQueryOutputFormat, which performs one BigQuery job for each Hadoop/Spark task.
See the example here: https://cloud.google.com/dataproc/docs/tutorials/bigquery-connector-spark-example
Related
What is the Best performance architecture to read XML in Spring Batch? Each XML is approximately 300 KB size and we are processing 1 Million.
Our Current Approach
30 partitions and 30 Grids and Each slave gets 166 XMLS
Commit Chunk 100
Application Start Memory is 8 GB
Using JAXB in Reader Default Bean Scope
#StepScope
#Qualifier("xmlItemReader")
public IteratorItemReader<BaseDTO> xmlItemReader(
#Value("#{stepExecutionContext['fileName']}") List<String> fileNameList) throws Exception {
String readingFile = "File Not Found";
logger.info("----StaxEventItemReader----fileName--->" + fileNameList.toString());
List<BaseDTO> fileList = new ArrayList<BaseDTO>();
for (String filePath : fileNameList) {
try {
readingFile = filePath.trim();
Invoice bill = (Invoice) getUnMarshaller().unmarshal(new File(filePath));
UnifiedInvoiceDTO unifiedDTO = new UnifiedInvoiceDTO(bill, environment);
unifiedDTO.setFileName(filePath);
BaseDTO baseDTO = new BaseDTO();
baseDTO.setUnifiedDTO(unifiedDTO);
fileList.add(baseDTO);
} catch (Exception e) {
UnifiedInvoiceDTO unifiedDTO = new UnifiedInvoiceDTO();
unifiedDTO.setFileName(readingFile);
unifiedDTO.setErrorMessage(e);
BaseDTO baseDTO = new BaseDTO();
baseDTO.setUnifiedDTO(unifiedDTO);
fileList.add(baseDTO);
}
}
return new IteratorItemReader<>(fileList);
}
Our questions:
Is this Archirecture correct
Is any performance or architecture advantage of using StaxEventItemReader and XStreamMarshaller over JAXB.
How to handle memory properly to avoid slow down
I would create a job per xml file by using the file name as a job parameter. This approach has many benefits:
Restartability: If a job fails, you only restart the failed file (from where it left off)
Scalability: This approach allows you to run multiple jobs in parallel. If a single machine is not enough, you can distribute the load on multiple machines
Logging: Logs are separate by design, you don't need to use an MDC or any other technique to separate logs
We are receiving XML filepath in a *.txt file
You can a create a script that iterates over these lines and launch a job per line (aka per file). Gnu Parallel (or a similar tool) is a good option to launch jobs in parallel.
I'm new to using Storm, I've just started a Data Architect training course and it's in this context that I'm facing the problem that brings me to you today.
I'm receiving messages from kakfa via a KafkaSpout named CurrentPriceSpout. So far, everything is working. Then, in my CurrentPriceBolt, I re-issued a tuple so that my data is written in ElasticSearch using the EsCurrentPriceBolt . The problem is here. I can't write my data into ElasticSearch directly, it is only written when I delete my topology.
Is there a Storm parameter that can force the writing of tuples by retrieving acknowledgments?
I tried by adding the option ".addConfiguration(Config.TOPOLOGY_TICK_TUPLE_FREQ_SECS, 5)", the tuples are well written in ElasticSearch but not acknowledged. So Storm rewrites them indefinitely.
Thanks for your help
Thierry
I managed to find the answer to my problem.
The main problem was that ES is not designed to ingest as little data as is generated in a study project. ES writes, by default, data in batches of 1000 entries. With this project, I generate one data every 30 seconds, or a batch of 1000 every 500 minutes (or 8h20).
so I reviewed in detail the configuration of my topology and played with the following options:
es.batch.size.entries: 1
es.storm.bolt.flush.entries.size: 1
topology.producer.batch.size: 1
topology.transfer.batch.size: 1
And now it goes like this:
...
...
public class App
{
...
...
public static void main( String[] args ) throws AlreadyAliveException, InvalidTopologyException, AuthorizationException
{
...
...
StormTopology topology = topologyBuilder.createTopology(); // je crée ma topologie Storm
String topologyName = properties.getProperty("storm.topology.name"); // je nomme ma topologie
StormSubmitter.submitTopology(topologyName, getTopologyConfig(properties), topology); // je démarre ma topologie sur mon cluster storm
System.out.println( "Topology on remote cluster : Started!" );
}
private static Config getTopologyConfig(Properties properties)
{
Config stormConfig = new Config();
stormConfig.put("topology.workers", Integer.parseInt(properties.getProperty("topology.workers")));
stormConfig.put("topology.enable.message.timeouts", Boolean.parseBoolean(properties.getProperty("topology.enable.message.timeouts")));
stormConfig.put("topology.message.timeout.secs", Integer.parseInt(properties.getProperty("topology.message.timeout.secs")));
stormConfig.put("topology.transfer.batch.size", Integer.parseInt(properties.getProperty("topology.transfer.batch.size")));
stormConfig.put("topology.producer.batch.size", Integer.parseInt(properties.getProperty("topology.producer.batch.size")));
return stormConfig;
}
...
...
...
}
And now it works!!!
We are receiving data in spark streaming from Kafka. Once execution has been started in Spark Streaming, it executes only one batch and the remaining batches starting queuing up in Kafka.
Our data is independent and can be processes in Parallel.
We tried multiple configurations with multiple executor, cores, back pressure and other configurations but nothing worked so far. There are a lot messages queued and only one micro batch has been processed at a time and rest are remained in queue.
We want to achieve parallelism at maximum, so that not any micro batch is queued, as we have enough resources available. So how we can reduce time by maximum utilization of resources.
// Start reading messages from Kafka and get DStream
final JavaInputDStream<ConsumerRecord<String, byte[]>> consumerStream = KafkaUtils.createDirectStream(
getJavaStreamingContext(), LocationStrategies.PreferConsistent(),
ConsumerStrategies.<String, byte[]>Subscribe("TOPIC_NAME",
sparkServiceConf.getKafkaConsumeParams()));
ThreadContext.put(Constants.CommonLiterals.LOGGER_UID_VAR, CommonUtils.loggerUniqueId());
JavaDStream<byte[]> messagesStream = consumerStream.map(new Function<ConsumerRecord<String, byte[]>, byte[]>() {
private static final long serialVersionUID = 1L;
#Override
public byte[] call(ConsumerRecord<String, byte[]> kafkaRecord) throws Exception {
return kafkaRecord.value();
}
});
// Decode each binary message and generate JSON array
JavaDStream<String> decodedStream = messagesStream.map(new Function<byte[], String>() {
private static final long serialVersionUID = 1L;
#Override
public String call(byte[] asn1Data) throws Exception {
if(asn1Data.length > 0) {
try (InputStream inputStream = new ByteArrayInputStream(asn1Data);
Writer writer = new StringWriter(); ) {
ByteArrayInputStream byteArrayInputStream = new ByteArrayInputStream(asn1Data);
GZIPInputStream gzipInputStream = new GZIPInputStream(byteArrayInputStream);
byte[] buffer = new byte[1024];
ByteArrayOutputStream byteArrayOutputStream = new ByteArrayOutputStream();
int len;
while((len = gzipInputStream.read(buffer)) != -1) {
byteArrayOutputStream.write(buffer, 0, len);
}
return new String(byteArrayOutputStream.toByteArray());
} catch (Exception e) {
//
producer.flush();
throw e;
}
}
return null;
}
});
// publish generated json gzip to kafka
cache.foreachRDD(new VoidFunction<JavaRDD<String>>() {
private static final long serialVersionUID = 1L;
#Override
public void call(JavaRDD<String> jsonRdd4DF) throws Exception {
//Dataset<Row> json = sparkSession.read().json(jsonRdd4DF);
if(!jsonRdd4DF.isEmpty()) {
//JavaRDD<String> jsonRddDF = getJavaSparkContext().parallelize(jsonRdd4DF.collect());
Dataset<Row> json = sparkSession.read().json(jsonRdd4DF);
SparkAIRMainJsonProcessor airMainJsonProcessor = new SparkAIRMainJsonProcessor();
airMainJsonProcessor.processAIRData(json, sparkSession);
}
}
});
getJavaStreamingContext().start();
getJavaStreamingContext().awaitTermination();
getJavaStreamingContext().stop();
Technology that we are using:
HDFS 2.7.1.2.5
YARN + MapReduce2 2.7.1.2.5
ZooKeeper 3.4.6.2.5
Ambari Infra 0.1.0
Ambari Metrics 0.1.0
Kafka 0.10.0.2.5
Knox 0.9.0.2.5
Ranger 0.6.0.2.5
Ranger KMS 0.6.0.2.5
SmartSense 1.3.0.0-1
Spark2 2.0.x.2.5
Statistics that we got from difference experimentations:
Experiment 1
num_executors=6
executor_memory=8g
executor_cores=12
100 Files processing time 48 Minutes
Experiment 2
spark.default.parallelism=12
num_executors=6
executor_memory=8g
executor_cores=12
100 Files processing time 8 Minutes
Experiment 3
spark.default.parallelism=12
num_executors=6
executor_memory=8g
executor_cores=12
100 Files processing time 7 Minutes
Experiment 4
spark.default.parallelism=16
num_executors=6
executor_memory=8g
executor_cores=12
100 Files processing time 10 Minutes
Please advise, how we can process maximum so no queued.
I was facing same issue and I tried a few things in trying to resolve the issue and came to following findings:
First of all. Intuition says that one batch must be processed per executor but on the contrary, only one batch is processed at a time but jobs and tasks are processed in parallel.
Multiple batch processing can be achieved by using spark.streaming.concurrentjobs, but it's not documented and still needs a few fixes. One of problems is with saving Kafka offsets. Suppose we set this parameter to 4 and 4 batches are processed in parallel, what if 3rd batch finishes before 4th one, which Kafka offsets would be committed. This parameter is quite useful if batches are independent.
spark.default.parallelism because of its name is sometimes considered to make things parallel. But its true benefit is in distributed shuffle operations. Try different numbers and find an optimum number for this. You will get a considerable difference in processing time. It depends upon shuffle operations in your jobs. Setting it too high would decrease the performance. It's apparent from you experiments results too.
Another option is to use foreachPartitionAsync in place of foreach on RDD. But I think foreachPartition is better as foreachPartitionAsync would queue up the jobs whereas batches would appear to be processed but their jobs would still be in the queue or in processing. May be I didn't get its usage right. But it behaved same in my 3 services.
FAIR spark.scheduler.mode must be used for jobs with lots of tasks as round-robin assignment of tasks to jobs, gives opportunity to smaller tasks to start receiving resources while bigger tasks are processing.
Try to tune your batch duration+input size and always keep it below processing duration otherwise you're gonna see a long backlog of batches.
These are my findings and suggestions, however, there are so many configurations and methods to do streaming and often one set of operation doesn't work for others. Spark Streaming is all about learning, putting your experience and anticipation together to get to a set of optimum configuration.
Hope it helps. It would be a great relief if someone could tell specifically how we can legitimately process batches in parallel.
We want to achieve parallelism at maximum, so that not any micro batch is queued
That's the thing about stream processing: you process the data in the order it was received. If you process your data at the rate slower than it arrives it will be queued. Also, don't expect that processing of one record will suddenly be parallelized across multiple nodes.
From your screenshot, it seems your batch time is 10 seconds and your producer published 100 records over 90 seconds.
It took 36s to process 2 records and 70s to process 17 records. Clearly, there is some per-batch overhead. If this dependency is linear, it would take only 4:18 to process all 100 records in a single mini-batch thus beating your record holder.
Since your code is not complete, it's hard to tell what exactly takes so much time. Transformations in the code look fine but probably the action (or subsequent transformations) are the real bottlenecks. Also, what's with producer.flush() which wasn't mentioned anywhere in your code?
I was facing the same issue and I solved it using Scala Futures.
Here are some link that show how to use it:
https://alvinalexander.com/scala/how-use-multiple-scala-futures-in-for-comprehension-loop
https://www.beyondthelines.net/computing/scala-future-and-execution-context/
Also, this is piece of my code when I used Scala Futures:
messages.foreachRDD{ rdd =>
val f = Future {
// sleep(100)
val newRDD = rdd.map{message =>
val req_message = message.value()
(message.value())
}
println("Request messages: " + newRDD.count())
var resultrows = newRDD.collect()//.collectAsList()
processMessage(resultrows, mlFeatures: MLFeatures, conf)
println("Inside scala future")
1
}
f.onComplete {
case Success(messages) => println("yay!")
case Failure(exception) => println("On no!")
}
}
It's hard to tell without having all the details, but general advice to tackle issues like that -- start with very simple application, "Hello world" kind. Just read from input stream and print data into log file. Once this works you prove that problem was in application and you gradually add your functionality back until you find what was culprit. If even simplest app doesn't work - you know that problem in configuration or Spark cluster itself. Hope this helps.
I am running a Spark Streaming application based on mapWithState DStream function . The application transforms input records into sessions based on a session ID field inside the records.
A session is simply all of the records with the same ID . Then I perform some analytics on a session level to find an anomaly score.
I couldn't stabilize my application because a handful of sessions are getting bigger at each batch time for extended period ( more than 1h) . My understanding is a single session (key - value pair) is always processed by a single core in spark . I want to know if I am mistaken , and if there is a solution to mitigate this issue and make the streaming application stable.
I am using Hadoop 2.7.2 and Spark 1.6.1 on Yarn . Changing batch time, blocking interval , partitions number, executor number and executor resources didn't solve the issue as one single task makes the application always choke. However, filtering those super long sessions solved the issue.
Below is a code updateState function I am using :
val updateState = (batchTime: Time, key: String, value: Option[scala.collection.Map[String,Any]], state: State[Seq[scala.collection.Map[String,Any]]]) => {
val session = Seq(value.getOrElse(scala.collection.Map[String,Any]())) ++ state.getOption.getOrElse(Seq[scala.collection.Map[String,Any]]())
if (state.isTimingOut()) {
Option(null)
} else {
state.update(session)
Some((key,value,session))
}
}
and the mapWithStae call :
def updateStreamingState(inputDstream:DStream[scala.collection.Map[String,Any]]): DStream[(String,Option[scala.collection.Map[String,Any]], Seq[scala.collection.Map[String,Any]])] ={//MapWithStateDStream[(String,Option[scala.collection.Map[String,Any]], Seq[scala.collection.Map[String,Any]])] = {
val spec = StateSpec.function(updateState)
spec.timeout(Duration(sessionTimeout))
spec.numPartitions(192)
inputDstream.map(ds => (ds(sessionizationFieldName).toString, ds)).mapWithState(spec)
}
Finally I am applying a feature computing session foreach DStream , as defined below :
def computeSessionFeatures(sessionId:String,sessionRecords: Seq[scala.collection.Map[String,Any]]): Session = {
val features = Functions.getSessionFeatures(sessionizationFeatures,recordFeatures,sessionRecords)
val resultSession = new Session(sessionId,sessionizationFieldName,sessionRecords)
resultSession.features = features
return resultSession
}
There is a fair amount of info online about bulk loading to HBase with Spark streaming using Scala (these two were particularly useful) and some info for Java, but there seems to be a lack of info for doing it with PySpark. So my questions are:
How can data be bulk loaded into HBase using PySpark?
Most examples in any language only show a single column per row being upserted. How can I upsert multiple columns per row?
The code I currently have is as follows:
if __name__ == "__main__":
context = SparkContext(appName="PythonHBaseBulkLoader")
streamingContext = StreamingContext(context, 5)
stream = streamingContext.textFileStream("file:///test/input");
stream.foreachRDD(bulk_load)
streamingContext.start()
streamingContext.awaitTermination()
What I need help with is the bulk load function
def bulk_load(rdd):
#???
I've made some progress previously, with many and various errors (as documented here and here)
So after much trial and error, I present here the best I have come up with. It works well, and successfully bulk loads data (using Puts or HFiles) I am perfectly willing to believe that it is not the best method, so any comments/other answers are welcome. This assume you're using a CSV for your data.
Bulk loading with Puts
By far the easiest way to bulk load, this simply creates a Put request for each cell in the CSV and queues them up to HBase.
def bulk_load(rdd):
#Your configuration will likely be different. Insert your own quorum and parent node and table name
conf = {"hbase.zookeeper.qourum": "localhost:2181",\
"zookeeper.znode.parent": "/hbase-unsecure",\
"hbase.mapred.outputtable": "Test",\
"mapreduce.outputformat.class": "org.apache.hadoop.hbase.mapreduce.TableOutputFormat",\
"mapreduce.job.output.key.class": "org.apache.hadoop.hbase.io.ImmutableBytesWritable",\
"mapreduce.job.output.value.class": "org.apache.hadoop.io.Writable"}
keyConv = "org.apache.spark.examples.pythonconverters.StringToImmutableBytesWritableConverter"
valueConv = "org.apache.spark.examples.pythonconverters.StringListToPutConverter"
load_rdd = rdd.flatMap(lambda line: line.split("\n"))\#Split the input into individual lines
.flatMap(csv_to_key_value)#Convert the CSV line to key value pairs
load_rdd.saveAsNewAPIHadoopDataset(conf=conf,keyConverter=keyConv,valueConverter=valueConv)
The function csv_to_key_value is where the magic happens:
def csv_to_key_value(row):
cols = row.split(",")#Split on commas.
#Each cell is a tuple of (key, [key, column-family, column-descriptor, value])
#Works well for n>=1 columns
result = ((cols[0], [cols[0], "f1", "c1", cols[1]]),
(cols[0], [cols[0], "f2", "c2", cols[2]]),
(cols[0], [cols[0], "f3", "c3", cols[3]]))
return result
The value converter we defined earlier will convert these tuples into HBase Puts
Bulk loading with HFiles
Bulk loading with HFiles is more efficient: rather than a Put request for each cell, an HFile is written directly and the RegionServer is simply told to point to the new HFile. This will use Py4J, so before the Python code we have to write a small Java program:
import py4j.GatewayServer;
import org.apache.hadoop.hbase.*;
public class GatewayApplication {
public static void main(String[] args)
{
GatewayApplication app = new GatewayApplication();
GatewayServer server = new GatewayServer(app);
server.start();
}
}
Compile this, and run it. Leave it running as long as your streaming is happening. Now update bulk_load as follows:
def bulk_load(rdd):
#The output class changes, everything else stays
conf = {"hbase.zookeeper.qourum": "localhost:2181",\
"zookeeper.znode.parent": "/hbase-unsecure",\
"hbase.mapred.outputtable": "Test",\
"mapreduce.outputformat.class": "org.apache.hadoop.hbase.mapreduce.HFileOutputFormat2",\
"mapreduce.job.output.key.class": "org.apache.hadoop.hbase.io.ImmutableBytesWritable",\
"mapreduce.job.output.value.class": "org.apache.hadoop.io.Writable"}#"org.apache.hadoop.hbase.client.Put"}
keyConv = "org.apache.spark.examples.pythonconverters.StringToImmutableBytesWritableConverter"
valueConv = "org.apache.spark.examples.pythonconverters.StringListToPutConverter"
load_rdd = rdd.flatMap(lambda line: line.split("\n"))\
.flatMap(csv_to_key_value)\
.sortByKey(True)
#Don't process empty RDDs
if not load_rdd.isEmpty():
#saveAsNewAPIHadoopDataset changes to saveAsNewAPIHadoopFile
load_rdd.saveAsNewAPIHadoopFile("file:///tmp/hfiles" + startTime,
"org.apache.hadoop.hbase.mapreduce.HFileOutputFormat2",
conf=conf,
keyConverter=keyConv,
valueConverter=valueConv)
#The file has now been written, but HBase doesn't know about it
#Get a link to Py4J
gateway = JavaGateway()
#Convert conf to a fully fledged Configuration type
config = dict_to_conf(conf)
#Set up our HTable
htable = gateway.jvm.org.apache.hadoop.hbase.client.HTable(config, "Test")
#Set up our path
path = gateway.jvm.org.apache.hadoop.fs.Path("/tmp/hfiles" + startTime)
#Get a bulk loader
loader = gateway.jvm.org.apache.hadoop.hbase.mapreduce.LoadIncrementalHFiles(config)
#Load the HFile
loader.doBulkLoad(path, htable)
else:
print("Nothing to process")
Finally, the fairly straightforward dict_to_conf:
def dict_to_conf(conf):
gateway = JavaGateway()
config = gateway.jvm.org.apache.hadoop.conf.Configuration()
keys = conf.keys()
vals = conf.values()
for i in range(len(keys)):
config.set(keys[i], vals[i])
return config
As you can see, bulk loading with HFiles is more complex than using Puts, but depending on your data load it is probably worth it since once you get it working it's not that difficult.
One last note on something that caught me off guard: HFiles expect the data they receive to be written in lexical order. This is not always guaranteed to be true, especially since "10" < "9". If you have designed your key to be unique, then this can be fixed easily:
load_rdd = rdd.flatMap(lambda line: line.split("\n"))\
.flatMap(csv_to_key_value)\
.sortByKey(True)#Sort in ascending order