David, I'm sure your version of Riak CS is 1.5.0. At 1.5.0 we fixed bug #475 by traversing whole bucket in creation and deletion [1]. This traverse reads all objects in the bucket, even if it is already deleted. Deleted objects remains physically for a while - (called leeway seconds) - with it's manifests. Checking bucket before deletion or creation reads manifests in a scheduled_delete state. That's why it's slow. If you wait for leeway seconds and make sure GC cleaned up all of them, bucket creation or deletion will be as fast as before creating many objects.
[2] is a little bit deeper description of how object chunking and garbage collection work, for further reading. [1] https://github.com/basho/riak_cs/blob/release/1.5/RELEASE-NOTES.md#incomplete-multipart-uploads [2] https://github.com/basho/riak_cs/wiki/Object-Chunking-and-Garbage-Collection On Fri, Aug 15, 2014 at 5:12 AM, David Whittington <[email protected]> wrote: > In testing Riak CS I've noticed that buckets that have had a lot of object > updates take a while to delete even if they're empty at the time of > deletion. Also, after deletion, creating a new bucket with the same name > takes a similar amount of time. This effect seems to persist indefinitely. > > Here's an example: https://gist.github.com/djwhitt/9480a1adb52a4b87d246 > > While this isn't a huge problem since the operation ultimately succeeds and > it doesn't need to be super fast for my use case, it does make me curious. > What is going on here? Is something not getting cleaned up properly? Anyone > else experiencing anything like this? > > -- > David Whittington > > _______________________________________________ > riak-users mailing list > [email protected] > http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com > -- Kota UENISHI / @kuenishi Basho Japan KK _______________________________________________ riak-users mailing list [email protected] http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
