Recently I've encountered an issue running one of our PySpark jobs. While analyzing the stages in Spark UI I have noticed that the longest running stage takes 1.2 hours to run out of the total 2.5 hours that takes for the entire process to run.
Once I took a look at the stage details it was clear that I'm facing a severe data skew, causing a single task to run for the entire 1.2 hours while all other tasks finish within 23 seconds.
The DAG showed this stage involves Window Functions which helped me to quickly narrow down the problematic area to a few queries and finding the root cause -> The column, account, that was being used in the Window.partitionBy("account") had 25% of null values.
I don't have an interest to calculate the sum for the null accounts though I do need the involved rows for further calculations therefore I can't filter them out prior the window function.
Here is my window function query:
problematic_account_window = Window.partitionBy("account")
sales_with_account_total_df = sales_df.withColumn("sum_sales_per_account", sum(col("price")).over(problematic_account_window))
So we found the one to blame - What can we do now? How can we resolve the skew and the performance issue?
We basically have 2 solutions for this issue:
Break the initial dataframe to 2 different dataframes, one that filters out the null values and calculates the sum on, and the second that contains only the null values and is not part of the calculation. Lastly we union the two together.
Apply salting technique on the null values in order to spread the nulls on all partitions and provide stability to the stage.
Solution 1:
account_window = Window.partitionBy("account")
# split to null and non null
non_null_accounts_df = sales_df.where(col("account").isNotNull())
only_null_accounts_df = sales_df.where(col("account").isNull())
# calculate the sum for the non null
sales_with_non_null_accounts_df = non_null_accounts_df.withColumn("sum_sales_per_account", sum(col("price")).over(account_window)
# union the calculated result and the non null df to the final result
sales_with_account_total_df = sales_with_non_null_accounts_df.unionByName(only_null_accounts_df, allowMissingColumns=True)
Solution 2:
SPARK_SHUFFLE_PARTITIONS = spark.conf.get("spark.sql.shuffle.partitions")
modified_sales_df = (sales_df
# create a random partition value that spans as much as number of shuffle partitions
.withColumn("random_salt_partition", lit(ceil(rand() * SPARK_SHUFFLE_PARTITIONS)))
# use the random partition values only in case the account value is null
.withColumn("salted_account", coalesce(col("account"), col("random_salt_partition")))
)
# modify the partition to use the salted account
salted_account_window = Window.partitionBy("salted_account")
# use the salted account window to calculate the sum of sales
sales_with_account_total_df = sales_df.withColumn("sum_sales_per_account", sum(col("price")).over(salted_account_window))
In my solution I've decided to use solution 2 since it didn't force me to create more dataframes for the sake of the calculation, and here is the result:
As seen above the salting technique helped resolving the skewness. The exact same stage now runs for a total of 5.5 minutes instead of 1.2 hours. The only modification in the code was the salting column in the partitionBy. The comparison shown is based on the exact same cluster/nodes amount/cluster config.
Related
I've got a problem with Spark, its driver and an OoM issue.
Currently I have a dataframe which is being built with several, joined sources (actually different tables in parquet format), and there are thousands of tuples. They have a date which represents the date of creation of the record, and distinctly they are a few.
I do the following:
from pyspark.sql.functions import year, month
# ...
selectionRows = inputDataframe.select(year('registration_date').alias('year'), month('registration_date').alias('month')).distinct()
selectionRows.show() # correctly shows 8 tuples
selectionRows = selectionRows.collect() # goes heap space OoM
print(selectionRows)
Reading the memory consumption statistics shows that the driver does not exceed ~60%. I thought that the driver should load only the distinct subset, not the entire dataframe.
Am I missing something? Is it possible to collect those few rows in a smarter way? I need them as a pushdown predicate to load a secondary dataframe.
Thank you very much!
EDIT / SOLUTION
After reading the comments and elaborating my personal needs, I cached the dataframe at every "join/elaborate" step, so that in a timeline I do the following:
Join with loaded table
Queue required transformations
Apply the cache transformation
Print the count to keep track of cardinality (mainly for tracking / debugging purposes) and thus apply all transformations + cache
Unpersist the cache of the previous sibiling step, if available (tick/tock paradigm)
This reduced some complex ETL jobs down to 20% of the original time (as previously it was applying the transformations of each previous step at each count).
Lesson learned :)
After reading the comments, I elaborated the solution for my use case.
As mentioned in the question, I join several tables one with each other in a "target dataframe", and at each iteration I do some transformations, like so:
# n-th table work
target = target.join(other, how='left')
target = target.filter(...)
target = target.withColumn('a', 'b')
target = target.select(...)
print(f'Target after table "other": {target.count()}')
The problem of slowliness / OoM was that Spark was forced to do all the transformations from start to finish at each table due to the ending count, making it slower and slower at each table / iteration.
The solution I found is to cache the dataframe at each iteration, like so:
cache: DataFrame = null
# ...
# n-th table work
target = target.join(other, how='left')
target = target.filter(...)
target = target.withColumn('a', 'b')
target = target.select(...)
target = target.cache()
target_count = target.count() # actually do the cache
if cache:
cache.unpersist() # free the memory from the old cache version
cache = target
print(f'Target after table "other": {target_count}')
I am working on Spark SQL where I need to find out Diff between two large CSV's.
Diff should give:-
Inserted Rows or new Record // Comparing only Id's
Changed Rows (Not include inserted ones) - Comparing all column values
Deleted rows // Comparing only Id's
Spark 2.4.4 + Java
I am using Databricks to Read/Write CSV
Dataset<Row> insertedDf = newDf_temp.join(oldDf_temp,oldDf_temp.col(key)
.equalTo(newDf_temp.col(key)),"left_anti");
Long insertedCount = insertedDf.count();
logger.info("Inserted File Count == "+insertedCount);
Dataset<Row> deletedDf = oldDf_temp.join(newDf_temp,oldDf_temp.col(key)
.equalTo(newDf_temp.col(key)),"left_anti")
.select(oldDf_temp.col(key));
Long deletedCount = deletedDf.count();
logger.info("deleted File Count == "+deletedCount);
Dataset<Row> changedDf = newDf_temp.exceptAll(oldDf_temp); // This gives rows (New +changed Records)
Dataset<Row> changedDfTemp = changedDf.join(insertedDf, changedDf.col(key)
.equalTo(insertedDf.col(key)),"left_anti"); // This gives only changed record
Long changedCount = changedDfTemp.count();
logger.info("Changed File Count == "+changedCount);
This works well for CSV with columns upto 50 or so.
The Above code fails for one row in CSV with 300+columns, so I am sure this is not file Size problem.
But if I have a CSV having 300+ Columns then it fails with Exception
Max iterations (100) reached for batch Resolution – Spark Error
If I set the below property in Spark, It Works!!!
sparkConf.set("spark.sql.optimizer.maxIterations", "500");
But my question is why do I have to set this?
Is there something wrong which I am doing?
Or this behaviour is expected for CSV's which have large columns.
Can I optimize it in any way to handle Large column CSV's.
The issue you are running into is related to how spark takes the instructions you tell it and transforms that into the actual things it's going to do. It first needs to understand your instructions by running Analyzer, then it tries to improve them by running its optimizer. The setting appears to apply to both.
Specifically your code is bombing out during a step in the Analyzer. The analyzer is responsible for figuring out when you refer to things what things you are actually referring to. For example, mapping function names to implementations or mapping column names across renames, and different transforms. It does this in multiple passes resolving additional things each pass, then checking again to see if it can resolve move.
I think what is happening for your case is each pass probably resolves one column, but 100 passes isn't enough to resolve all of the columns. By increasing it you are giving it enough passes to be able to get entirely through your plan. This is definitely a red flag for a potential performance issue, but if your code is working then you can probably just increase the value and not worry about it.
If it isn't working, then you will probably need to try to do something to reduce the number of columns used in your plan. Maybe combining all the columns into one encoded string column as the key. You might benefit from checkpointing the data before doing the join so you can shorten your plan.
EDIT:
Also, I would refactor your above code so you could do it all with only one join. This should be a lot faster, and might solve your other problem.
Each join leads to a shuffle (data being sent between compute nodes) which adds time to your job. Instead of computing adds, deletes and changes independently, you can just do them all at once. Something like the below code. It's in scala psuedo code because I'm more familiar with that than the Java APIs.
import org.apache.spark.sql.functions._
var oldDf = ..
var newDf = ..
val changeCols = newDf.columns.filter(_ != "id").map(col)
// Make the columns you want to compare into a single struct column for easier comparison
newDf = newDF.select($"id", struct(changeCols:_*) as "compare_new")
oldDf = oldDF.select($"id", struct(changeCols:_*) as "compare_old")
// Outer join on ID
val combined = oldDF.join(newDf, Seq("id"), "outer")
// Figure out status of each based upon presence of old/new
// IF old side is missing, must be an ADD
// IF new side is missing, must be a DELETE
// IF both sides present but different, it's a CHANGE
// ELSE it's NOCHANGE
val status = when($"compare_new".isNull, lit("add")).
when($"compare_old".isNull, lit("delete")).
when($"$compare_new" != $"compare_old", lit("change")).
otherwise(lit("nochange"))
val labeled = combined.select($"id", status)
At this point, we have every ID labeled ADD/DELETE/CHANGE/NOCHANGE so we can just a groupBy/count. This agg can be done almost entirely map side so it will be a lot faster than a join.
labeled.groupBy("status").count.show
Following function is supposed to join two DataFrames and return the number of checkouts per location. It is based on the Seattle Public Library data set.
def topKCheckoutLocations(checkoutDF: DataFrame, libraryInventoryDF: DataFrame, k: Int): DataFrame = {
checkoutDF
.join(libraryInventoryDF, "ItemType")
.groupBy("ItemBarCode", "ItemLocation") //grouping by ItemBarCode and ItemLocation
.agg(count("ItemBarCode")) //counting number of ItemBarCode for each ItemLocation
.withColumnRenamed("count(ItemBarCode)", "NumCheckoutItemsAtLocation")
.select($"ItemLocation", $"NumCheckoutItemsAtLocation")
}
When I run this, it takes ages to finish (40+ minutes), and I'm pretty sure it is not supposed to take more than a couple of minutes. Can I change the order of the calls to decrease computation time?
As I never managed to finish computation I never actually got to check whether the output is correct. I assume it is.
The checkoutDF has 3 mio. rows.
For spark job performance
Select the required column from the dataset before joins to
decrease data size
Partition your both dataset by join column ("ItemType") to avoid shuffling
I have a Spark DataFrame where all fields are integer type. I need to count how many individual cells are greater than 0.
I am running locally and have a DataFrame with 17,000 rows and 450 columns.
I have tried two methods, both yielding slow results:
Version 1:
(for (c <- df.columns) yield df.where(s"$c > 0").count).sum
Version 2:
df.columns.map(c => df.filter(df(c) > 0).count)
This calculation takes 80 seconds of wall clock time. With Python Pandas, it takes a fraction of second. I am aware that for small data sets and local operation, Python may perform better, but this seems extreme.
Trying to make a Spark-to-Spark comparison, I find that running MLlib's PCA algorithm on the same data (converted to a RowMatrix) takes less than 2 seconds!
Is there a more efficient implementation I should be using?
If not, how is the seemingly much more complex PCA calculation so much faster?
What to do
import org.apache.spark.sql.functions.{col, count, when}
df.select(df.columns map (c => count(when(col(c) > 0, 1)) as c): _*)
Why
Your both attempts create number of jobs proportional to the number of columns. Computing the execution plan and scheduling the job alone are expensive and add significant overhead depending on the amount of data.
Furthermore, data might be loaded from disk and / or parsed each time the job is executed, unless data is fully cached with significant memory safety margin which ensures that the cached data will not be evicted.
This means that in the worst case scenario nested-loop-like structure you use can roughly quadratic in terms of the number of columns.
The code shown above handles all columns at the same time, requiring only a single data scan.
The problem with your approach is that the file is scanned for every column (unless you have cached it in memory). The fastet way with a single FileScan should be:
import org.apache.spark.sql.functions.{explode,array}
val cnt: Long = df
.select(
explode(
array(df.columns.head,df.columns.tail:_*)
).as("cell")
)
.where($"cell">0).count
Still I think it will be slower than with Pandas, as Spark has a certain overhead due to the parallelization engine
I have an ML dataframe which I read from csv files. It contains three types of columns:
ID Timestamp Feature1 Feature2...Feature_n
where n is ~ 500 (500 features in ML parlance). The total number of rows in the dataset is ~ 160 millions.
As this is the result of a previous full join, there are many features which do not have values set.
My aim is to run a "fill" function(fillna style form python pandas), where each empty feature value gets set with the previously available value for that column, per Id and Date.
I am trying to achieve this with the following spark 2.2.1 code:
val rawDataset = sparkSession.read.option("header", "true").csv(inputLocation)
val window = Window.partitionBy("ID").orderBy("DATE").rowsBetween(-50000, -1)
val columns = Array(...) //first 30 columns initially, just to see it working
val rawDataSetFilled = columns.foldLeft(rawDataset) { (originalDF, columnToFill) =>
originalDF.withColumn(columnToFill, coalesce(col(columnToFill), last(col(columnToFill), ignoreNulls = true).over(window)))
}
I am running this job on a 4 m4.large instances on Amazon EMR, with spark 2.2.1. and dynamic allocation enabled.
The job runs for over 2h without completing.
Am I doing something wrong, at the code level? Given the size of the data, and the instances, I would assume it should finish in a reasonable amount of time? And I haven't even tried with the full 500 columns, just with about 30!
Looking in the container logs, all I see are many logs like this:
INFO codegen.CodeGenerator: Code generated in 166.677493 ms
INFO execution.ExternalAppendOnlyUnsafeRowArray: Reached spill
threshold of
4096 rows, switching to
org.apache.spark.util.collection.unsafe.sort.UnsafeExternalSorter
I have tried setting parameter spark.sql.windowExec.buffer.spill.threshold to something larger, without any impact. Is theresome other setting I should know about? Those 2 lines are the only ones I see in any container log.
In Ganglia, I see most of the CPU cores peaking around full usage, but the memory usage is lower than the maximum available. All executors are allocated and are doing work.
I have managed to rewrite the fold left logic without using withColumn calls. Apparently they can be very slow for large number of columns, and I was also getting stackoverflow errors because of that.
I would be curious to know why this massive difference - and what exactly happens behind the scenes with the query plan execution, which makes repeated withColumns calls so slow.
Links which proved very helpful: Spark Jira issue and this stackoverflow question
var rawDataset = sparkSession.read.option("header", "true").csv(inputLocation)
val window = Window.partitionBy("ID").orderBy("DATE").rowsBetween(Window.unboundedPreceding, Window.currentRow)
rawDataset = rawDataset.select(rawDataset.columns.map(column => coalesce(col(column), last(col(column), ignoreNulls = true).over(window)).alias(column)): _*)
rawDataset.write.option("header", "true").csv(outputLocation)