Details
-
Bug
-
Status: Resolved
-
Critical
-
Resolution: Won't Fix
-
Private Beta
-
None
-
None
Description
I ran a script to delete all the tables on the YCSB cluster (~50GB or so per host I think, so not even a huge amount). This ended up taking quite some time – the DeleteTablet RPC seems to synchronously try to delete all of the blocks, which can be quite slow (we do an fsync per block apparently, vs some kind of smarter batching). Seems like we should be able to do this all asynchronously.