cshannon commented on PR #3640: URL: https://github.com/apache/accumulo/pull/3640#issuecomment-1693997778
@keith-turner - You can ignore that original comment about the DeleteRowsIT hanging. It turns out the client was getting exceptions with trying to load Tablets that were offline. Things got screwed up as metadata was being updated for online tablets but that has been fixed. I have deletes (seemingly) working without chop compactions in a branch that is based off of this PR: https://github.com/cshannon/accumulo/tree/no-chop-delete It loads the tablets that overlap the start/end of the delete range and then if the tablets share a boundary with the delete range (end row matches prevs end row) then the files are fenced. This is the same condition used to detect when to run a chop compaction. This isn't final because the condition on when metadata is updated is likely going to change when I look at getting rid of splits which I'll do next. Currently this works because splits run first and create the new tablets on the deletion boundaries. When splits go away in favor of fencing the files, i'll need to update the condition on when fencing happens but i at least wanted to get this working for now. At some point (when I work on splits) I guess I'll want to create a PR for the combination of no chop deletes and fencing files instead of splits but that runs into the chained PR issue so we will see how that goes, but until then I'll keep it just in a branch. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
