We have a table of 130GB and 4000 columns. When we select 2 of these columns, our Spark UI reports a total of 30GB read. However, if we select those two columns and store them as a separate dataset, the total size of the dataset is just 17MB. Given that parquet is columnar storage, something appears not to be working properly. I've found this question but I'm unsure how to diagnose further and what attempts to take to reduce the amount of I/O required.
It was my understanding that the benefit of columnar storage is that each column can be read more or less independently of each other.
We're running on Hadoop 2.7.X on Databricks. It occurs both on the 6.X and 7.X versions of databricks (spark 2.4/3.0)
Related
We have a requirement to ingest data from a non-partitioned EXTERNAL hive table work_db.customer_tbl to a partitioned EXTERNAL hive table final_db.customer_tbl through PySpark, previously done through hive query. The final table is partitioned by the column load_date (format of load_date column is yyyy-MM-dd).
So we have a simple PySpark script which uses an insert query (same as the hive query which was used earlier), to ingest the data using spark.sql() command. But we have some serious performance issues because the table we are trying to ingest after ingestion has around 3000 partitions and each partitions has around 4 MB of data except for the last partition which is around 4GB. Total table size is nearly 15GB. Also, after ingestion each partition has 217 files. The final table is a snappy compressed parquet table.
The source work table has a single 15 GB file with filename in the format customers_tbl_unload.dat.
Earlier when we were using the hive query through a beeline connection it usually takes around 25-30 minutes to finish. Now when we are trying to use the PySpark script it is taking around 3 hours to finish.
How can we tune the spark performance to make the ingestion time less than what it took for beeline.
The configurations of the yarn queue we use is:
Used Resources: <memory:5117184, vCores:627>
Demand Resources: <memory:5120000, vCores:1000>
AM Used Resources: <memory:163072, vCores:45>
AM Max Resources: <memory:2560000, vCores:500>
Num Active Applications: 45
Num Pending Applications: 45
Min Resources: <memory:0, vCores:0>
Max Resources: <memory:5120000, vCores:1000>
Reserved Resources: <memory:0, vCores:0>
Max Running Applications: 200
Steady Fair Share: <memory:5120000, vCores:474>
Instantaneous Fair Share: <memory:5120000, vCores:1000>
Preemptable: true
The parameters passed to the PySpark script is:
num-executors=50
executor-cores=5
executor-memory=10GB
PySpark code used:
insert_stmt = """INSERT INTO final_db.customers_tbl PARTITION(load_date)
SELECT col_1,col_2,...,load_date FROM work_db.customer_tbl"""
spark.sql(insert_stmt)
Even after nearly using 10% resources of the yarn queue the job is taking so much time. How can we tune the job to make it more efficient.
You need to reanalyze your dataset and look if you are using the correct approach by partitioning yoir dataset on date column or should you be probably partitioning on year?
To understand why you end up with 200 plus files for each partition, you need to understand the difference between the Spark and Hive partitions.
A direct approach you should try first is to read your input dataset as a dataframe and partition it by the key you are planning to use as a partition key in Hive and then save it using df.write.partitionBy
Since the data seems to be skewed too on date column, try partitioning it on additional columns which might have equal distribution of data. Else, filter out the skewed data and process it separately
I am working on a pipeline that will run daily. It includes joining 2 tables say x & y ( approx. 18 MB and 1.5 GB sizes respectively) and loading the output of the join to final table.
Following are the facts about the environment,
For table x:
Data size: 18 MB
Number of files in a partition : ~191
file type: parquet
For table y:
Data size: 1.5 GB
Number of files in a partition : ~3200
file type: parquet
Now the problem is:
Hive and Spark are giving same performance (time taken is same)
I tried different combination of resources for spark job.
e.g.:
executors:50 memory:20GB cores:5
executors:70 memory:20GB cores:5
executors:1 memory:20GB cores:5
All three combinations are giving same performance. I am not sure what I am missing here.
I also tried broadcasting the small table 'x' so as to avoid shuffle while joining but not much improvement in performance.
One key observations is:
70% of the execution time is consumed for reading the big table 'y' and I guess this is due to more number of files per partition.
I am not sure how hive is giving the same performance.
Kindly suggest.
I assume you are comparing Hive on MR vs Spark. Please let me know if it is not the case.Because Hive(on tez or spark) vs Spark Sql will not differ
vastly in terms of performance.
I think the main issue is that there are too many small files.
A lot of CPU and time is consumed in the I/O itself, hence you can't experience the processing power of Spark.
My advice is to coalesce the spark dataframes immedietely after reading the parquet files. Please coalesce the 'x' dataframe into single partition and 'y'
dataframe into 6-7 partitions.
After doing the above, please perform the join(broadcastHashJoin).
What is the difference between append and overwrite to parquet in spark.
I'm processing huge amount of data for say 10 days. At present I'm processing daily logs into parquet files using "append" method and partitioning the data based on date. But the problem I'm facing is daily data is also very huge and taking a lot of time, contributing to high CPU usage as well while processing data using EMR cluster. This is making my job very slow and expensive. So I'm looking for a way where I can further split the data and can merge the data to day cluster.
Please see spark SaveMode docs
https://spark.apache.org/docs/latest/api/java/index.html
I use Spark 2.
Actually I am not the one executing the queries so I cannot include query plans. I have been asked this question by the data science team.
We are having hive table partitioned into 2000 partitions and stored in parquet format. When this respective table is used in spark, there are exactly 2000 tasks that are executed among the executors. But we have a block size of 256 MB and we are expecting the (total size/256) number of partitions which will be much lesser than 2000 for sure. Is there any internal logic that spark uses physical structure of data to create partitions. Any reference/help would be greatly appreciated.
UPDATE: It is the other way around. Actually our table is very huge like 3 TB having 2000 partitions. 3TB/256MB would actually come to 11720 but we are having exactly same number of partitions as the table is partitioned physically. I just want to understand how the tasks are generated on data volume.
In general Hive partitions are not mapped 1:1 to Spark partitions. 1 Hive partition can be split into multiple Spark partitions, and one Spark partition can hold multiple hive-partitions.
The number of Spark partitions when you load a hive-table depends on the parameters:
spark.files.maxPartitionBytes (default 128MB)
spark.files.openCostInBytes (default 4MB)
You can check the partitions e.g. using
spark.table(yourtable).rdd.partitions
This will give you an Array of FilePartitions which contain the physical path of your files.
Why you got exactly 2000 Spark partitions from your 2000 hive partitions seems a coincidence to me, in my experience this is very unlikely to happen. Note that the situation in spark 1.6 was different, there the number of spark partitions resembled the number of files on the filesystem (1 spark partition for 1 file, unless the file was very large)
I just want to understand how the tasks are generated on data volume.
Tasks are a runtime artifact and their number is exactly the number of partitions.
The number of tasks does not correlate to data volume in any way. It's a Spark developer's responsibility to have enough partitions to hold the data.
I am working with a large dataset, that is partitioned by two columns - plant_name and tag_id. The second partition - tag_id has 200000 unique values, and I mostly access the data by specific tag_id values. If I use the following Spark commands:
sqlContext.setConf("spark.sql.hive.metastorePartitionPruning", "true")
sqlContext.setConf("spark.sql.parquet.filterPushdown", "true")
val df = sqlContext.sql("select * from tag_data where plant_name='PLANT01' and tag_id='1000'")
I would expect a fast response as this resolves to a single partition. In Hive and Presto this takes seconds, however in Spark it runs for hours.
The actual data is held in a S3 bucket, and when I submit the sql query, Spark goes off and first gets all the partitions from the Hive metastore (200000 of them), and then calls refresh() to force a full status list of all these files in the S3 object store (actually calling listLeafFilesInParallel).
It is these two operations that are so expensive, are there any settings that can get Spark to prune the partitions earlier - either during the call to the metadata store, or immediately afterwards?
Yes, spark supports partition pruning.
Spark does a listing of partitions directories (sequential or parallel listLeafFilesInParallel) to build a cache of all partitions first time around. The queries in the same application, that scan data takes advantage of this cache. So the slowness that you see could be because of this cache building. The subsequent queries that scan data make use of the cache to prune partitions.
These are the logs which shows partitions being listed to populate the cache.
App > 16/11/14 10:45:24 main INFO ParquetRelation: Listing s3://test-bucket/test_parquet_pruning/month=2015-01 on driver
App > 16/11/14 10:45:24 main INFO ParquetRelation: Listing s3://test-bucket/test_parquet_pruning/month=2015-02 on driver
App > 16/11/14 10:45:24 main INFO ParquetRelation: Listing s3://test-bucket/test_parquet_pruning/month=2015-03 on driver
These are the logs showing pruning is happening.
App > 16/11/10 12:29:16 main INFO DataSourceStrategy: Selected 1 partitions out of 20, pruned 95.0% partitions.
Refer convertToParquetRelation and getHiveQlPartitions in HiveMetastoreCatalog.scala.
Just a thought:
Spark API documentation for HadoopFsRelation says,
( https://spark.apache.org/docs/1.6.2/api/java/org/apache/spark/sql/sources/HadoopFsRelation.html )
"...when reading from Hive style partitioned tables stored in file
systems, it's able to discover partitioning information from the paths
of input directories, and perform partition pruning before start
reading the data..."
So, i guess "listLeafFilesInParallel" could not be a problem.
A similar issue is already in spark jira: https://issues.apache.org/jira/browse/SPARK-10673
In spite of "spark.sql.hive.verifyPartitionPath" set to false and, there is no effect in performance, I suspect that the
issue might have been caused by unregistered partitions. Please list out the partitions of the table and verify if all
the partitions are registered. Else, recover your partitions as shown in this link:
Hive doesn't read partitioned parquet files generated by Spark
Update:
I guess appropriate parquet block size and page size were set while writing the data.
Create a fresh hive table with partitions mentioned, and file-format as parquet, load it from non-partitioned table using dynamic partition approach.
( https://cwiki.apache.org/confluence/display/Hive/DynamicPartitions )
Run a plain hive query and then compare by running a spark program.
Disclaimer: I am not a spark/parquet expert. The problem sounded interesting, and hence responded.
similar question popped up here recently:
http://apache-spark-user-list.1001560.n3.nabble.com/Spark-SQL-reads-all-leaf-directories-on-a-partitioned-Hive-table-td35997.html#a36007
This question is old but I thought I'd post the solution here as well.
spark.sql.hive.convertMetastoreParquet=false
will use the Hive parquet serde instead of the spark inbuilt parquet serde. Hive's Parquet serde will not do a listLeafFiles on all partitions, but only and directly read from the selected partitions. On tables with many partitions and files, this is much faster (and cheaper, too). Feel free to try it ou! :)