On Fri, 10 May 2013, Noah Watkins wrote: > Mike, > > Thanks for the looking into this further. > > On May 10, 2013, at 5:23 AM, Mike Bryant <mike.bry...@ocado.com> wrote: > > > I've just found this bug report though: http://tracker.ceph.com/issues/3601 > > Looks like that may be the same issue.. > > This definitely seems like a candidate.
Yeah. Although in this case, there is only one client with caps, and it is excl, which makes me think it's not quite the same as 3601. Client needs to make a concerted effort to push unlinked inodes out of its own cache, and it doesn't currently do this. We probably need to put the unlinked inode (or its parent dentries) on a special list so that it gets forced out of cache on the next trim. sage > > >> Adding some debug to the cephfs jar, I was able to match up all opens with > >> either a current existing file in cephfs, or a matching close. > > There is already debugging present in the Java bindings. You can turn on > client logging, and add 'debug javaclient = 20' to get client debug logs. > > How many clients does HBase setup? > _______________________________________________ > ceph-users mailing list > ceph-users@lists.ceph.com > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > > _______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
