Closed russss closed 7 years ago
@JeremyGrosser I've rebased this on top of the current master - would you be able to take a look at it?
We've been running this code successfully for a few years now, and it definitely adds quite a significant speedup to tablechop when you've got tens of thousands of sstables. Now I'm familiar with the tablesnap code I'm tempted to make a few more PRs, but it would be cool to see this one merged first.
This parallelises tablechop's fetching of the .json files to reduce the amount of time it takes to run. It adds a dependency on eventlet.
This also removes the
keep_existing
option in favour of a different approach - it wasn't really clear to me what the intention ofkeep_existing
was:Instead of keeping all files which exist on disk, plus all json files which reference it, this update will always keep the last backup.