Ceph » RADOS

One might face an assert on collection's release which happens
after store destroy. For now is observable in some qa/standalone test cases in debug build only:

2018-11-27 13:18:15.788 7fe8b47b8380 -1 ** Caught signal (Aborted) *
in thread 7fe8b47b8380 thread_name:ceph-osd

ceph version 14.0.1-1055-ge3f828c97a (e3f828c97a185d74a0d31a52ab5a8635c0ef17fa) nautilus (dev)
 1: (()+0x2b1db44) [0x55e19b4a7b44]
 2: (()+0x11f70) [0x7fe8b3497f70]
 3: (gsignal()+0x10b) [0x7fe8b2b8b08b]
 4: (abort()+0x123) [0x7fe8b2b744e9]
 5: (()+0x94973) [0x7fe8b3354973]
 6: (()+0x9a8c8) [0x7fe8b335a8c8]
 7: (()+0x999b9) [0x7fe8b33599b9]
 8: (_gxx_personality_v0()+0x2c5) [0x7fe8b335a305]
 9: (()+0x10da3) [0x7fe8b2f23da3]
 10: (_Unwind_RaiseException()+0x2b1) [0x7fe8b2f24301]
 11: (_cxa_throw()+0x36) [0x7fe8b335ab36]
 12: (ceph::mutex_debug_detail::mutex_debug_impl<true>::try_lock_impl()+0x7c) [0x55e19b35e728]
 13: (ceph::mutex_debug_detail::mutex_debug_impl<true>::try_lock(bool)+0x1d) [0x55e19b3475fb]
 14: (ceph::mutex_debug_detail::mutex_debug_impl<true>::lock(bool)+0x4c) [0x55e19b332496]
 15: (std::lock_guard<ceph::mutex_debug_detail::mutex_debug_impl<true> >::lock_guard(ceph::mutex_debug_detail::mutex_debug_impl<true>&)+0x2f) [0x55e19b31e63d]
 16: (BlueStore::OnodeSpace::clear()+0x3f) [0x55e19b27fae5]
 17: (BlueStore::OnodeSpace::~OnodeSpace()+0x18) [0x55e19b315f32]
 18: (BlueStore::Collection::~Collection()+0x42) [0x55e19b3831ac]
 19: (BlueStore::Collection::~Collection()+0x18) [0x55e19b383208]
 20: (RefCountedObject::put() const+0x2b7) [0x55e19ab3a2fb]
 21: (()+0x2132fcd) [0x55e19aabcfcd]
 22: (boost::intrusive_ptr<ObjectStore::CollectionImpl>::~intrusive_ptr()+0x27) [0x55e19ab63f15]
 23: (OSD::mkfs(CephContext*, ObjectStore*, uuid_d, int)+0xede) [0x55e19aad084a]