Perhaps you had a DDL collision and ended up with two data dirs for the table?
In that case running drop table would only move the active table directory to snapshots and as Eric suggested would leave the data in the duplicate directory "orphaned". I haven't tried to reproduce this yet but I think given how DDL works in c* it checks out as a possible scenario. Keep calm and Cassandra on folks, Seb On Thu, Apr 30, 2020, 7:46 PM Sergio <lapostadiser...@gmail.com> wrote: > The problem is that folder is not under snapshot but it is under the data > path. > I tried with the --all switch too > Thanks, > Sergio > > On Thu, Apr 30, 2020, 4:21 PM Nitan Kainth <nitankai...@gmail.com> wrote: > >> I don't think it works like that. clearsnapshot --all would remove all >> snapshots. Here is an example: >> >> $ ls -l >> /ss/xx/cassandra/data/ww/a-5bf825428b3811eabe0c6b7631a60bb0/snapshots/ >> >> total 8 >> >> drwxr-xr-x 2 cassandra cassandra 4096 Apr 30 23:17 >> dropped-1588288650821-a >> >> drwxr-xr-x 2 cassandra cassandra 4096 Apr 30 23:17 manual >> >> $ nodetool clearsnapshot --all >> >> Requested clearing snapshot(s) for [all keyspaces] with [all snapshots] >> >> $ ls -l >> /ss/xx/cassandra/data/ww/a-5bf825428b3811eabe0c6b7631a60bb0/snapshots/ >> >> ls: cannot access >> /ss/xx/cassandra/data/ww/a-5bf825428b3811eabe0c6b7631a60bb0/snapshots/: >> No such file or directory >> >> $ >> >> >> On Thu, Apr 30, 2020 at 5:44 PM Erick Ramirez <erick.rami...@datastax.com> >> wrote: >> >>> Yes, you're right. It doesn't show up in listsnapshots nor does >>> clearsnapshot remove the dropped snapshot because the table is no >>> longer managed by C* (because it got dropped). So you will need to manually >>> remove the dropped-* directories from the filesystem. >>> >>> Someone here will either correct me or hopefully provide a >>> user-friendlier solution. Cheers! >>> >>
