Re:Running Fast SQL on DynamoDB Tables
by iamsomewalrus on Wednesday January 23, 2019 @07:24PM
(#18984763)
I’ll have a blog post out in AWS Big Data in a few weeks but the parent is right in that it’s not terribly difficult.
Should take about an hour with testing to get a Pyspark script together to read in a DynamoDB table and write it out to S3.
https://docs.aws.amazon.com/glue/latest/dg/aws-glue-programm...
You’ll then need to crawl the S3 data to add it to your Glue catalog and then you can query it with Athena.
Re:Running Fast SQL on DynamoDB Tables
by alexbilbie on Friday January 25, 2019 @08:19AM
(#18997634)
If you're running this regularly (e.g. once an hour) to dump tables the cost of Glue can really add up due a minimum runtime length.
We switched to a scheduled Fargate task to dump data from DynamoDB into S3 as parquet files. It's really reliable, costs us ~$4/month and completely configurable.
I ask because I have looked and haven't found one that I could follow without having to dedicate several hours to it.
To the point where I want to move off of dynamodb to aurora because I can't believe how hard it is to get simple insights like how many rows exist out of dynamo (I inherited the dynamo choice would not have used it myself)