I'm looking for a way to set-up an incremental Glue crawler for S3 data, where data arrives continuously and is partitioned by the date it was captured (so the S3 paths within the include path contain date=yyyy-mm-dd). My concern is, that if I run the crawler in the course of a day, the partition for it will be created, and will not be re-visited in subsequent crawls. Is there a way to force a given partition, that I know might still be receiving updates, to be crawled while running the crawler incrementally and not wasting resources on historic data?
The crawler will visit only new folders with an incremental crawl (assuming you have set crawl new folders only option). The only circumstance where adding more data to an existing folder would cause a problem is if you were changing schema by adding a differently formatted file into a folder that was already crawled. Otherwise the crawler has created the partition and knows the schema, and is ready to pull the data, even if new files are added to the existing folder.