I have to migrate to Spark 2.1 an application written in Scala 2.10.4 using Spark 1.6.
The application treats text files with around 7GB of dimension, and contains several rdd transformations.
I was told to try to recompile it with scala 2.11, which should be enough to make it work with Spark 2.1. This sounds strange to me as I know in Spark 2 there are some relevant changes, like:
Introduction of SparkSession object
Merge of DataSet and DataFrame
APIs
I managed to recompile the application in spark 2 with scala 2.11 with only minor changes due to Kryo Serializer registration.
I still have some runtime error that I am trying to solve and I am trying to figure out what will come next.
My question regards what changes are "neccessary" in order to make the application work as before, and what changes are "recommended" in terms of performance optimization (I need to keep at least the same level of performances), and whatever you think could be useful for a newbie in spark :).
Thanks in advance!
I did the same 1 year ago, there are not many changes you need to do, what comes in my mind:
if your code is cluttered with spark/sqlContext, then just extract this variable from SparkSession instace at the beginning of your code.
df.map switched to RDD API in Spark 1.6, in Spark 2.+ you stay in DataFrame API (which now has a map method). To get same functionality as before, replace df.map with df.rdd.map. The same is true for df.foreach and df.mapPartitions etc
unionAll in Spark 1.6 is just union in Spark 2.+
The databrick csv library is now included in Spark.
When you insert into a partitioned hive table, then the partition columns must now come as last column in the schema, in Spark 1.6 it had to be the first column
What you should consider (but would require more work):
migrate RDD-Code into Dataset-Code
enable CBO (cost based optimizer)
collect_list can be used with structs, in Spark 1.6 it could only be used with primitives. This can simplify some things
Datasource API was improved/unified
leftanti join was introduced
Related
I am trying to do a bulk insert or bulk load into Hbase in EMR using Glue Scala (Spark 3.1). I got this using
table.put(List<Put>);
without a satisfatory performance. I tried to insert though spark dataframe following some examples but the libraries features are compatible just with Spark 1.6. I tried, too, reproduce some examples of insert HFiles into HDFS enviroment and processing through HOutputFormat and HOutputFormat2 but these classes were removed from newer versions. How can I be able to perform a highly-performatic insert in HBase, with current libraries or, even, an bulkload? The examples that I found were old and the Hbase Book Reference wasn't clearly about this point.
Thank you.
I'm quite new to pySpark but I'm confused about the difference between a spark Dataframe (created for example from an RDD ) and a pandas-on-spark Dataframe.
Are those the same object ? Looking at the type it seems they are different classes.
What's the core difference, if any ? (I know that working with pandas-on-spark Dataframe you can use almost the same syntax of Pandas on a distributed Dataframe but I'm wondering if is only this one the difference )
Thanks
Answering directly:
Are those the same object ? Looking at the type it seems they are different classes.
No, they are completely different objects (classes).
What's the core difference, if any ?
A pySpark DataFrame is an object from the PySpark library, with its own API and it can be constructed from a wide array of sources such as: structured data files, tables in Hive, external databases, or existing RDDs.
A Pandas-on-Spark DataFrame and pandas DataFrame are similar. However, the former is distributed and the latter is in a single machine. When converting to each other, the data is transferred between multiple machines and the single client machine.
A Pandas DataFrame, is an object from the pandas library, also with its own API and it can also be constructed from a wide range of methods.
Also, I recommend checking this documentation about Pandas on Spark
I am using Apache Spark DataFrame and I want to Delete data to Elasticsearch.
For adding and updating I am using below command
val esURL = "https://56h874526b6741db87c3c91324g755.westeurope.azure.elastic-cloud.com:9243"
var indexName = "test_elastic/test_elastic"
df.write
.format("org.elasticsearch.spark.sql")
.option("es.nodes.wan.only","true")
.option("es.port","443")
.option("es.net.ssl","true")
.option("es.net.http.auth.user","userid")
.option("es.net.http.auth.pass","pwd")
.option("es.nodes", esURL)
.option("es.mapping.id", "primary_key")
.mode("append")
.save(indexName)
My question is how I can delete some rows from elasticsearch index. In my case elasticsearch index is "test_elastic".
A quick search in the repository mentions first support for deletes on the ElasticSearch Hadoop connector for the 8.x and 7.8 versions, neither of which are currently released the moment I write this.
https://github.com/elastic/elasticsearch-hadoop/pull/1324
From maintainer jbaiera :
jbaiera left a comment •
LGTM! Thanks very much for your dedication on getting this in! I'll go ahead and merge it in to master and backport it to the 7.x branch. It should be available in the 7.8.0 release when that lands.
From the current code changes, supported ES versions should be 2.x to 8.x, but there is no documentation yet that I could find (did not look much though), and no information about a direct usage in the Spark API (and I'm not even sure a delete API exists on Spark Dataframes, whatever their data source).
On the other hand, there exists a write mode called "overwrite", that is working and could allow you achieve data deletion, but overwriting the whole index may not be practical depending on the volumetry.
I think your best bet would be to drop out of the spark dataframe API to switch to one (or several) direct call(s) to bulk delete.
I want to do a full Scan on hbase from Spark 2 using Scala.
I don't have a fixed catalog definition so libraries as SHC are not an option.
My logical choice was to use hbase-spark, that is working fine in Spark 1.6
In addition to the poor documentation about this library in previous versions, my surprise has been when checking the last HBase releasees, for example tag 2.0, hbase-spark is gone! but still in the master.
So my questions are:
Where is the hbase-spark module for the last releases?
Where can I find a hbase-spark version compatible with Spark 2?
thx!
Seems hbase-spark module was removed from the hbase project for v2.0 release
https://issues.apache.org/jira/browse/HBASE-18817
#bp2010 already answered part of the question.
Regarding the HBase Spark see below. It works with spark 2.
There are some options that don't demand a fixed catalog from client code:
HBase Spark
Source code with examples are here: https://github.com/apache/hbase-connectors/tree/master/spark/hbase-spark
Here you can see explanations about the repositories:
https://github.com/apache/hbase-connectors/tree/master/spark/hbase-spark
Apache Phoenix Spark connector
https://phoenix.apache.org/phoenix_spark.html
I'm not sure if it helps you, since the table must be mapped to a Phoenix table. If you have Phoenix, and you problem is writing the catalog from code, but you can standardize types in HBase Table, for a full scan this can be the way to go. Otherwise, go with option 1.
I am new to Scala, and i have to use Scala and Spark's SQL, Mllib and GraphX in order to perform some analysis on huge data set. The analyses i want to do are:
Customer life cycle Value (CLV)
Centrality measures (degree, Eigenvector, edge-betweenness,
closeness) The data is in a CSV file (60GB (3 years transnational data))
located in Hadoop cluster.
My question is about the optimal approach to access the data and perform the above calculations?
Should i load the data from the CSV file into dataframe and work on
the dataframe? or
Should i load the data from the CSV file and convert it into RDD and
then work on the RDD? or
Are there any other approach to access the data and perform the analyses?
Thank you so much in advance for your help..
Dataframe gives you sql like syntax to work with the data where as RDD gives Scala collection like methods for data manipulation.
One extra benefit with Dataframes is underlying spark system will optimise your queries just like sql query optimisation. This is not available in case of RDD's.
As you are new to Scala its highly recommended to use Dataframes API initially and then Pick up RDD API later based on requirement.
You can use Databricks CSV reader api, which is easy to use and returns DataFrame. It automatically infer data types. If you pass the file with header it can automatically use that as Schema, otherwise you can construct schema using StructType.
https://github.com/databricks/spark-csv
Update:
If you are using Spark 2.0 Version , by default it support CSV datasource, please see the below link.
https://spark.apache.org/releases/spark-release-2-0-0.html#new-features
See this link for how to use.
https://github.com/databricks/spark-csv/issues/367