I have created multiple parquet files using pyspark and now I'm trying to merge all the parquet files to 1. I'm able to merge the files, but while reading in the resulting file, I'm getting an error. Have anyone faced this issue before?
You cannot simply concatenate Parquet files using cat as they are binary files with a "table of contents" in the footer. To merge two files, you would have to read them both in and write out a completely new file. This could be done easily using the merge command in the parquet-tools.
The technical background that merging two Parquet files using cat isn't working comes down to the fact that a Parquet file is useless without a footer. Every Parquet file is made up roughly by the following structure:
RowGroup(nrows=..)
Column with nrows
Column with nrows
..
RowGroup(nrows=..)
..
..
Footer
Schema (tells you the type of the columns)
total_nrows
Location of RowGroups in the file
If you cat two files together, you would only see the last footer of the two files. Additionally, if the library used to read the files does some integrity checks, it will realise that your file is broken in some fashion and completely error out.
Related
I'm using SPARK to read files in hdfs. There is a scenario, where we are getting files as chunks from legacy system in csv format.
ID1_FILENAMEA_1.csv
ID1_FILENAMEA_2.csv
ID1_FILENAMEA_3.csv
ID1_FILENAMEA_4.csv
ID2_FILENAMEA_1.csv
ID2_FILENAMEA_2.csv
ID2_FILENAMEA_3.csv
This files are loaded to FILENAMEA in HIVE using HiveWareHouse Connector, with few transformation like adding default values. Similarly we have around 70 tables. Hive tables are created in ORC format. Tables are partitioned on ID. Right now, I'm processing all these files one by one. It's taking much time.
I want to make this process much faster. Files will be in GBs.
Is there is any way to read all the FILENAMEA files at the same time and load it to HIVE tables.
You have two methods to read several CSV files in pyspark. If all CSV files are in the same directory and all have the same schema, you can read then at once by directly passing the path of directory as argument, as follow:
spark.read.csv('hdfs://path/to/directory')
If you have CSV files in different locations or CSV files in same directory but with other CSV/text files in it, you can pass them as string representing a list of path in .csv() method argument, as follow:
spark.read.csv('hdfs://path/to/filename1,hdfs://path/to/filename2')
You can have more information about how to read a CSV file with Spark here
If you need to build this list of paths from the list of files in HDFS directory, you can look at this answer, once you've created your list of paths, you can transform it to a string to pass to .csv() method with ','.join(your_file_list)
Using: spark.read.csv(["path1","path2","path3"...]) you can read multiple files from different paths. But that means you have first to make a list of the paths. A list not a string of comma-separated file paths
Using .vacuum() on a DeltaLake table is very slow (see Delta Lake (OSS) Table on EMR and S3 - Vacuum takes a long time with no jobs).
If I manually deleted the underlying parquet files and did not add a new json log file or add a new .checkpoint.parquet file and change the _delta_log/_last_checkpoint file that points to it; what would the negative impacts to the DeltaLake table be, if any?
Obviously time-traveling, i.e. loading a previous version of the table that relied on the parquet files I removed, would not work. What I want to know is, would there be any issues reading, writing, or appending to the current version of the DeltaLake table?
What I am thinking of doing in pySpark:
### Assuming a working SparkSession as `spark`
from subprocess import check_output
import json
from pyspark.sql import functions as F
awscmd = "aws s3 cp s3://my_s3_bucket/delta/_delta_log/_last_checkpoint -"
last_checkpoint = str(json.loads(check_output(awscmd, shell=True).decode("utf-8")).get('version')).zfill(20)
s3_bucket_path = "s3a://my_s3_bucket/delta/"
df_chkpt_del = (
spark.read.format("parquet")
.load(f"{s3_bucket_path}/_delta_log/{last_checkpoint}.checkpoint.parquet")
.where(F.col("remove").isNotNull())
.select("remove.*")
.withColumn("deletionTimestamp", F.from_unixtime(F.col("deletionTimestamp")/1000))
.withColumn("delDateDiffDays", F.datediff(F.col("deletionTimestamp"), F.current_timestamp()))
.where(F.col("delDateDiffDays") < -7 )
)
There are a lot of options from here. One could be:
df_chkpt_del.select("path").toPandas().to_csv("files_to_delete.csv", index=False)
Where I could read files_to_delete.csv into a bash array and then use a simple bash for loop passing each parquet file s3 path to an aws s3 rm command to remove the files one by one.
This may be slower than vacuum(), but at least it will not be consuming cluster resources while it is working.
If I do this, will I also have to either:
write a new _delta_log/000000000000000#####.json file that correctly documents these changes?
write a new 000000000000000#####.checkpoint.parquet file that correctly documents these changes and change the _delta_log/_last_checkpoint file to point to that checkpoint.parquet file?
The second option would be easier.
However, if there will be no negative effects if I just remove the files and don't change anything in the _delta_log, then that would be the easiest.
TLDR. Answering this question.
If I manually deleted the underlying parquet files and did not add a new json log file or add a new .checkpoint.parquet file and change the _delta_log/_last_checkpoint file that points to it; what would the negative impacts to the DeltaLake table be, if any?
Yes, this could potentially corrupt your delta table.
Let me briefly answers how delta-lake reads a version using _delta_log.
If you want to read version x then it will go to delta log of all versions from 1 to x-1 and will make a running sum of parquet files to read. Summary of this process is saved as a .checkpoint after every 10th version to make this process of running sum efficient.
What do I mean by this running sum?
Assume,
version 1 log says, add add file_1, file_2, file_3
version 2 log says, add delete file_1, file_2, and add file_4
So when reading version no 2, total instruction will be
add file_1, file_2, file_3 -> delete file_1, file_2, and add file_4
So, resultant files read will be file_3 and file_4.
What if you delete a parquet from a file system?
Say in version 3, you delete file_4 from file system. If you don't use .vacuum then delta log will not know that file_4 is not present, it will try to read it and will fail.
I have a directory of CSV files. The files are named based on date similar to the image below:
I have many CSV files that go back to 2012.
So, I would like to read the CSV files that correspond to a certain date only. How is that could be possible in spark? In other words, I don't want my spark engine to bother and read all CSV files because my data is huge (TBs).
Any help is much appreciated!
You can specify a list of files to be processed when calling the load(paths) or csv(paths) methods from DataFrameReader.
So an option would be to list and filter files on the driver, then load only the "recent" files :
val files: Seq[String] = ???
spark.read.option("header","true").csv(files:_*)
Edit :
You can use this python code (not tested yet)
files=['foo','bar']
df=spark.read.csv(*files)
Scala 2.12 and Spark 2.2.1 here. I used the following code to write the contents of a DataFrame to S3:
myDF.write.mode(SaveMode.Overwrite)
.parquet("s3n://com.example.mybucket/mydata.parquet")
When I go to com.example.mybucket on S3 I actually see a directory called "mydata.parquet", as well as file called "mydata.parquet_$folder$"!!! If I go into the mydata.parquet directory I see two files under it:
_SUCCESS; and
part-<big-UUID>.snappy.parquet
Whereas I was just expecting to see a single file called mydata.parquet living in the root of the bucket.
Is something wrong here (if so, what?!?) or is this expected with the Parquet file format? If its expected, which is the actual Parquet file that I should read from:
mydata.parquet directory?; or
mydata.parquet_$folder$ file?; or
mydata.parquet/part-<big-UUID>.snappy.parquet?
Thanks!
The mydata.parquet/part-<big-UUID>.snappy.parquet is the actual parquet data file. However, often tools like Spark break data sets into multiple part files, and expect to be pointed to a directory that contains multiple files. The _SUCCESS file is a simple flag indicating that the write operation has completed.
According to the api to save the parqueat file it saves inside the folder you provide. Sucess is incidation that the process is completed scuesffuly.
S3 create those $folder if you write directly commit to s3. What happens is it writes to temporory folders and copies to the final destination inside the s3. The reason is there no concept of rename.
Look at the s3-distcp and also DirectCommiter for performance issue.
The $folder$ marker is used by s3n/amazon's emrfs to indicate "empty directory". ignore.
The _SUCCESS file is, as the others note, a 0-byte file. ignore
all other .parquet files in the directory are the output; the number you end up with depends on the number of tasks executed on the input
When spark uses a directory (tree) as a source of data, all files beginning with _ or . are ignored; s3n will strip out those $folder$ things too. So if you use the path for a new query, it will only pick up that parquet file.
I have a set of log files I would like to read into an RDD.
These files are all compressed .gz and are the filenames are date stamped.
The source of these files is the page view statistics data for wikipedia
http://dumps.wikimedia.org/other/pagecounts-raw/
The file names look like this:
pagecounts-20090501-000000.gz
pagecounts-20090501-010000.gz
pagecounts-20090501-020000.gz
What I would like to do is read in all such files in a directory and prepend the date from the filename (e.g. 20090501) to each row of the resulting RDD.
I first thought of using sc.wholeTextFiles(..) instead of sc.textFile(..), which creates a PairRDD with the key being the file name with a path,
but sc.wholeTextFiles() doesn't handle compressed .gz files.
Any suggestions would be welcome.
The following seems to work fine for me in Spark 1.6.0:
sc.wholeTextFiles("file:///tmp/*.gz").flatMapValues(y => y.split("\n")).take(10).foreach(println)
Sample output:
(file:/C:/tmp/pagecounts-20160101-000000.gz,aa 271_a.C 1 4675)
(file:/C:/tmp/pagecounts-20160101-000000.gz,aa Battaglia_di_Qade%C5%A1/it/Battaglia_dell%27Oronte 1 4765)
(file:/C:/tmp/pagecounts-20160101-000000.gz,aa Category:User_th 1
4770)
(file:/C:/tmp/pagecounts-20160101-000000.gz,aa Chiron_Elias_Krase 1 4694)