Ceph » CephFS

We need to be able to fix up the anchor table when there are problems, to avoid e.g.

(09:57:45 AM) failboat: mds/AnchorServer.cc: In function 'virtual void AnchorServer::handle_query(MMDSTableRequest*)':
(09:57:46 AM) failboat: mds/AnchorServer.cc:242: FAILED assert(anchor_map.count(curino) == 1)
(09:57:48 AM) failboat:  ceph version 0.23 (commit:5d1d8d0c4602be9819cc9f7aea562fccbb005a56)
(09:57:50 AM) failboat:  1: (AnchorServer::handle_query(MMDSTableRequest*)+0x7ed) [0x63c3ed]
(09:57:52 AM) failboat:  2: (MDS::_dispatch(Message*)+0x19a7) [0x4bdac7]
(09:57:54 AM) failboat:  3: (MDS::ms_dispatch(Message*)+0x5b) [0x4be79b]
(09:57:56 AM) failboat:  4: (SimpleMessenger::dispatch_entry()+0x6c9) [0x49ec79]
(09:57:58 AM) failboat:  5: (SimpleMessenger::DispatchThread::entry()+0x1c) [0x493bfc]
(09:58:00 AM) failboat:  6: (Thread::_entry_func(void*)+0xa) [0x4a689a]
(09:58:02 AM) failboat:  7: /lib64/libpthread.so.0() [0x397be06d5b]
(09:58:04 AM) failboat:  8: (clone()+0x6d) [0x397bae427d]
(09:58:06 AM) failboat: dunno if you've seen this
(09:58:11 AM) failboat: tell me if you need properly symbolized trace
(09:58:19 AM) failboat: this happens shortly after replay
(09:58:31 AM) failboat: when I access something