An automated data pipeline using Lambda, S3 and Glue - Big Data with Cloud Computing

[@parthakaizer]

Please keep uploading videos, you don't know how many lives you are touching man!

[@KnowledgeAmplifier1]

Thank you Partha Sarathi Barman, I will! Happy Learning :-)

[@radhikaa5462]

Very good informative tutorial, superb explanation of code and ETL pipeline ! I worked on Lambda before and this video helped me to work on Glue integrating with S3 and Lambda. Thank you !

[@KnowledgeAmplifier1]

Glad to know the video was helpful to you radhika a! Happy Learning :-)

[@Alexr26]

Awesome tutorial, I just used this information to create my first automated pipeline. Thanks a lot for sharing!

[@KnowledgeAmplifier1]

Congratulations Alejandro Aguirre😊! Happy Learning

[@parthakaizer]

I just had a requirement of this same solution. Thank you so much! You are a lifesaver!

[@KnowledgeAmplifier1]

Glad it helped Partha Sarathi Barman! Happy Learning :-)

[@XoXo-ou9yu]

@@KnowledgeAmplifier1 The script is creating Spark Session every time there is a file dropped in S3 bucket, This is making the Glue ETL extremally slow. How do we take care of that?

[@KnowledgeAmplifier1]

@@XoXo-ou9yu to avoid triggering 100 times when 100 files are uploaded, , designate one special key (file name) or prefix which will be used for trigger , once all the required files are written in s3 , then write the special file and configure the s3 event notification based on that file only , so that lambda will not trigger multiple times , it will trigger AWS Glue job only once (when the special file is written) .. You can check this video for details -- kzbin.info/www/bejne/paXQaIGYotusaac Hope this will be helpful! Happy Learning :-)

[@RajeshKumar-re8tj]

Hey, Please make course based on aws of 10-20 project availble for data engineer. I am intrested in your teaching style.

[@KnowledgeAmplifier1]

Hello Rajesh Kumar , Thank you for your interest in my teaching style. I'm delighted to hear that you're interested in an AWS-based data engineering course. I'd like to let you know that I've already created a comprehensive Medium blog which covers a range of projects for data engineers with in-depth explanations. You can access it to explore these projects and learn at your own pace : medium.com/@satadru1998/7-end-to-end-modern-data-engineering-projects-for-free-3c1c5f09d89e . If you have any questions or need further guidance, please feel free to reach out!

[@dianad150]

It looks really complicated to do a relatively simple job.

[@KnowledgeAmplifier1]

If Glue job has to be triggered based on s3 event, then as of now, Lambda is go-to solution, if there is any other easier work-around please share your thoughts 😊

[@gouthammadarapu8430]

Nice and clearly explained bro 👌

[@KnowledgeAmplifier1]

Thank you so much 🙂

[@datapragmatic]

Hi, why do you use a lambda for trigger the Glue's job? can't you use the Glue's triggers?

[@AliciaMarkoe]

Thank you 🦋

[@KnowledgeAmplifier1]

You are welcome Alicia Markoe! Happy Learning

[@sriadityab4794]

Thank you for the informative tutorial. Suppose if I have like files landing in different folders in S3 bucket with date partitions, how can we write the lambda code to trigger the glue job? Here I want the glue job to perform join operation for all the different sources files which I means until and unless all the files come and land in S3 in different folders with date partition, my glue job should not be triggered. Thanks

[@mayanktripathi4u]

Thanks for this video, its really helpful. Could you please also create a video on AWS DataPipeline may be for similar requirement which is accomplished in this video using AWS Glue. Actually I am facing an issue with AWS DataPipeline with custom scripts.

[@KnowledgeAmplifier1]

Sure Mayank Tripathi , will upload soon , Stay tuned , Happy Learning :-)

[@manojt7012]

@@KnowledgeAmplifier1 Hi bro. Can you make a video to differentiate when to use GLUE and when to use EMR. as both serves the similar purpose. Maybe Glue is high cost and serverless with less bootstraping time. Is there any other to consider?

[@KnowledgeAmplifier1]

@@manojt7012, we should use Glue for low/medium complexity ETL & EMR for high complexity ETL. If the data is some TB range data , if the transformation is not very heavy , if the process can complete within 1-2 hours & frequency of the job run is very high , then you can go with AWS Glue. If the ETL is having high complexity & the datasize is in Petabytes range (which is quite huge),if the job takes 6-7 hours to complete then EMR is preferred over Glue in this case (as it might happen that Glue may run out of memory but due to autoscaling of EMR it will not happen with EMR). Hope it will give some idea when to use EMR and when Glue. Happy Learning :-)

[@manojt7012]

@@KnowledgeAmplifier1 thanks a lot for clear explanation.

[@XoXo-ou9yu]

The script is creating Spark Session every time there is a file dropped in S3 bucket, This is making the Glue ETL extremally slow. How do we take care of that?

[@monamidatta1611]

Can I make lambda to trigger the glue job at a particular time? As I don't want to get it trigger when file is landing....as my files are landing the whole day...so recursion can happen...plz help

[@DiaaKasem0]

Thank you so much for this great video... But, the question that I have in mind, is for this use case you showed, wouldn't just using Lambda alone to make the transformation be sufficient ? .. no need to trigger Glue .. just read json from s3, flatten, write to output s3 ... or what is wrong with this idea ? .. and if you know, when what I am suggesting fails ? ... and when Glue will be really needed ? ( I expect in processing all existing files in a bucket ... or can we use lambda for that ? ) Thank You :) again

[@KnowledgeAmplifier1]

I am using spark for flattening the json ,and in Lambda , you can not run PySpark , so I used Glue , anyway if the file is small , you can flatten json with native python too but if the file is large enough to process within 15 mins , then you have to go with Glue as Lambda code run has max time limit of 15 mins!

[@DiaaKasem0]

@@KnowledgeAmplifier1 Thank You :) so much, I have that case where i have million of small files not big ones, so I always wondered why every were people used pyspark while lambda can do it, now I understand

[@XoXo-ou9yu]

@@DiaaKasem0 The script is creating Spark Session every time there is a file dropped in S3 bucket, This is making the Glue ETL extremally slow. How do we take care of that?

[@KnowledgeAmplifier1]

@@XoXo-ou9yu to avoid triggering 100 times when 100 files are uploaded, , designate one special key (file name) or prefix which will be used for trigger , once all the required files are written in s3 , then write the special file and configure the s3 event notification based on that file only , so that lambda will not trigger multiple times , it will trigger AWS Glue job only once (when the special file is written) .. You can check this video for details -- kzbin.info/www/bejne/paXQaIGYotusaac Hope this will be helpful! Happy Learning :-)

[@nmadhavirao]

@Knowledge Amplifier I have a question. I thought when we use Glue we have to crawl to create data catalog. In above example you did not do it. When do we crawl vs when do we not?

[@KnowledgeAmplifier1]

See glue crawler is optional stuff, if you are trying to infer schema from source system and then create a table in Glue Data Catalog and then want to use that table in Glue job then go with Crawler else if you are using normal spark code which can be used easily anywhere (onprem or EMR or glue wherever) then write a generalized code like I wrote and use glue as a platform to run light weight ETL 😊Hope you got this..and also remember for all the source system, you can't make connections with glue crawler and the source to crawl the data Happy Learning 😊

[@nmadhavirao]

@@KnowledgeAmplifier1 thank you so much for the reply. I have one more question. In real time, we will not be using AWS console to develop scripts. How would you write all code using python (including manual settings like like assigning AWS glue role etc that you did in the video with in the scripts) and how and where in AWS would you deploy the code?

[@aniruddhyadav4794]

@@nmadhavirao You can use cloud formation templates and code pipeline for real time without console

[@nmadhavirao]

@ Knowledge Amplifier Thank you for the video its very informative. I have one question: In real time, we will not be using AWS console to develop scripts. How would you write all code using python (including manual settings like like assigning AWS glue role etc that you did from scripts) and how and where in AWS would you deploy the code?

[@KnowledgeAmplifier1]

You can use AWS CLI or boto3

[@nmadhavirao]

@@KnowledgeAmplifier1 thank you

[@XoXo-ou9yu]

@@nmadhavirao The script is creating Spark Session every time there is a file dropped in S3 bucket, This is making the Glue ETL extremally slow. How do we take care of that?

[@KnowledgeAmplifier1]

@@XoXo-ou9yu to avoid triggering 100 times when 100 files are uploaded, , designate one special key (file name) or prefix which will be used for trigger , once all the required files are written in s3 , then write the special file and configure the s3 event notification based on that file only , so that lambda will not trigger multiple times , it will trigger AWS Glue job only once (when the special file is written) .. You can check this video for details -- kzbin.info/www/bejne/paXQaIGYotusaac Hope this will be helpful! Happy Learning :-)

[@durgarasane-kolapkar1842]

Thank for the tutorial Sir! In case, we get a burst of files in S3, there will be multiple invocations of Lambda and each lambda will also fire Glue Job. How to handle this scenario? Should we increase concurrency of Glue Job? If Yes - How much concurrency to set for Glue Job as we are not sure how many files will come to S3 at a time. Please advice

[@Arvindkumar-mb8yj]

Can ywe also load below CSV type file from S3 to snowflake ? There is CSV file with some comment in initial 2-3 line and then header starts and at the end there is one comment line which says total number of records in CSV file. How to crawl such CSV file and load into snowflake ?

[@XoXo-ou9yu]

Your script is creating Spark Session every time there is a file dropped in S3 bucket, This is making the Glue ETL extremally slow. How do we take care of that? @Knowledge Amplifier

[@KnowledgeAmplifier1]

to avoid triggering 100 times when 100 files are uploaded, , designate one special key (file name) or prefix which will be used for trigger , once all the required files are written in s3 , then write the special file and configure the s3 event notification based on that file only , so that lambda will not trigger multiple times , it will trigger AWS Glue job only once (when the special file is written) .. You can check this video for details -- kzbin.info/www/bejne/paXQaIGYotusaac Hope this will be helpful! Happy Learning :-)

[@XoXo-ou9yu]

@@KnowledgeAmplifier1 Thanks a lot!

[@XoXo-ou9yu]

@@KnowledgeAmplifier1 I used the script provided in the description, but it is only parsing the first line of the json file. and ignoring the rest of the file. Any idea why? using cloud watch I can see just a warn message "jndi lookup class is not available because this jre does not support jndi"

[@parthakaizer]

Just how do I write from Glue to a redshift table do you have any already created video?

[@KnowledgeAmplifier1]

You can check this -- docs.aws.amazon.com/glue/latest/dg/aws-glue-programming-etl-redshift.html Hope this will be helpful.. Happy Learning :-)

[@sagar7958]

Dhanyawad :)

[@KnowledgeAmplifier1]

No mention ...Happy Learning

[@niranjanjamkhande3773]

Great video. Thanks a lot. If the data is distributed in different folders in s3 or data lands in various folders in a single bucket, then how to modify codes and particular folder as trigger? Pls help. And I tried with nested json having variety of datatypes (int, struct, array). At that time could not get csv converted table in destination bucket. Its working only when data types are only array, and only struct.

[@sriadityab4794]

How to handle if there are multiple files dropped in S3 at the same time where we need trigger one glue job? How should we handle Lambda here? Any help is appreciated.

[@vikinist]

cant we use glue job(Data target) directly for converting json to csv?

[@sayalibendre4543]

what can be done if we want output file name not (part."""".....) in transformed bucket?

[@pinakisaha8179]

Thank you so much for this great video, I was able to execute it successfully, however in destination file its only adding 2 rows of data instead of all, what could be the reason? Do I need to modify the glue code? please advise.. Thanks again

[@saicharanpeddireddy5859]

how can we develop CI CD for this exact process.

[@KnowledgeAmplifier1]

Hello Saicharan Peddireddy 585, in my recent video , the process is explained , you can have a look if still looking for the answer -- kzbin.info/www/bejne/hGa9kIBnfbyHg6Msi=q_HG4aKkyln0zaM0 Hope this will be helpful! Happy Learning

[@krishnasanagavarapu4858]

awesome

[@KnowledgeAmplifier1]

You are welcome Krishna Sanagavarapu! Happy Learning & wish you a very Happy New Year :-)

[@krishnasanagavarapu4858]

@@KnowledgeAmplifier1

[@ardavanmoinzadeh801]

Where is the architecture diagram for this project?

[@KnowledgeAmplifier1]

Hello Ardavan Moinzadeh , you can have a look in this link -- github.com/SatadruMukherjee/Data-Preprocessing-Models/blob/main/AWS%20Glue%20Job%20trigger%20from%20Lambda.JPG Hope this will be helpful! Happy Learning

[@parthasarathibarman9862]

Hi Buddy, once again - one more question, I want to write clean production quality code - where from can I practice? any course? Also where from I can practice for data engineering projects using python. Any links to course free/paid which are genuinely good will really be appreciated. Thanks for your help in advance!

[@XoXo-ou9yu]

The script is creating Spark Session every time there is a file dropped in S3 bucket, This is making the Glue ETL extremally slow. How do we take care of that?

[@KnowledgeAmplifier1]

@@XoXo-ou9yu to avoid triggering 100 times when 100 files are uploaded, , designate one special key (file name) or prefix which will be used for trigger , once all the required files are written in s3 , then write the special file and configure the s3 event notification based on that file only , so that lambda will not trigger multiple times , it will trigger AWS Glue job only once (when the special file is written) .. You can check this video for details -- kzbin.info/www/bejne/paXQaIGYotusaac Hope this will be helpful! Happy Learning :-)

[@Truth___]

Very informative video 👍

[@KnowledgeAmplifier1]

Thank You Hritik bhai :-)

[@AA-gp2vv]

Thanks

[@KnowledgeAmplifier1]

You are Welcome AA ! Happy Learning :-)

[@sudipbala9647]

Thank you.

[@KnowledgeAmplifier1]

Welcoem Sudip Bala! Happy Learning :-)

[@veerachegu]

Any one here to help me i got project on aws glue

[@KnowledgeAmplifier1]

Can you please share your requirement

[@DanielWeikert]

I tried the same but converting to parquet from csv. I receive the following error Glue job 99.pyWriteDynamicFrame. Illegal empty schema any ideas? THanks

[@santoshmehta1999]

I am getting this error { "errorMessage": "'Records'", "errorType": "KeyError", "requestId": "caf276e6-da0a-44df-abdf-f62aec50c3ba", "stackTrace": [ " File \"/var/task/lambda_function.py\", line 6, in lambda_handler file_name = event['Records'][0]['s3']['object']['key'] " ] }

[@santoshmehta1999]

Got the solutuion

[@vaibhavverma1340]

@@santoshmehta1999 May I know how?????????????