Wissem, unfortunately there is no way to learn if zombies has appeared other than runnig fsck. But I think this can be perfomed on a weekly or even monthly basis - from my experience getting 32K zombies is a pretty rare case. But definitely it's more reliable if you collect that statistics from the cluster yourself... Thanks, Igor On 2/17/2022 5:43 PM, Wissem MIMOUNA wrote: Hi Igor, Thank you very much this helped us to understand the root cause and hope we will get a fix soon ( with new ceph release ) . In the means time do you have any idea how to perdiocally check the zombies spanning blobs ( before running the fsck/repair ) so it would be nice for us to automate this action ? Have a good day Best Regards De : Igor Fedotov <igor.fedotov@xxxxxxxx> Envoyé : jeudi 17 février 2022 11:59 À : Wissem MIMOUNA <wissem.mimouna@xxxxxxxxxxxxxxxx>; ceph-users@xxxxxxx Objet : Re: OSDs crash randomnisly Hi Wissem, first of all the bug wasn't fixed with the PR you're referring - it just added additional log output on the problem detection. Unfortunately the bug isn't fixed yet as the root cause for zombie spanning blobs appearance is still ZjQcmQRYFpfptBannerStart ZjQcmQRYFpfptBannerEnd Hi Wissem, first of all the bug wasn't fixed with the PR you're referring - it just added additional log output on the problem detection. Unfortunately the bug isn't fixed yet as the root cause for zombie spanning blobs appearance is still unclear. The relevant ticket is There is a workaround though - ceph-bluestore-tool's repair command would detect zombie spanning blobs and remove them which should eliminate the assertion for a while. I'd recommend to run fsck/repair periodically as it looks like your cluster is exposed to the problem and zombies would rather come back - it's crucial to keep their amount below 32K per PG to avoid the assertion. Thanks, Igor On 2/17/2022 1:41 PM, Wissem MIMOUNA wrote: > Dear, > > Some ODSs on our ceph cluster crush with no explication . > Stop and Start of the crushed OSD daemon fixed the issue but this happend few times and I just need to understand the reason. > For your information the error has been fixed in the log change in the octopus release ( > Below the logs related to the crash : > > > "process_name": "ceph-osd", > "entity_name": "osd.x", > "ceph_version": "15.2.15", > "utsname_hostname": "", > "utsname_sysname": "Linux", > "utsname_release": "4.15.0-162-generic", > "utsname_version": > "utsname_machine": "x86_64", > "os_name": "Ubuntu", > "os_id": "ubuntu", > "os_version_id": "18.04", > "os_version": "18.04.6 LTS (Bionic Beaver)", > "assert_condition": "abort", > "assert_func": "bid_t BlueStore::ExtentMap::allocate_spanning_blob_id()", > "assert_file": "/build/ceph-15.2.15/src/os/bluestore/BlueStore.cc", > "assert_line": 2664, > "assert_thread_name": "tp_osd_tp", > "assert_msg": "/build/ceph-15.2.15/src/os/bluestore/BlueStore.cc: In function 'bid_t BlueStore::ExtentMap::allocate_spanning_blob_id()' thread 7f6d37800700 time 2022-02-17T09:41:55.108101+0100\n/build/ceph-15.2.15/src/os/bluestore/BlueStore.cc: 2664: ceph_abort_msg(\"no available blob id\")\n", > "backtrace": [ > "(()+0x12980) [0x7f6d59516980]", > "(gsignal()+0xc7) [0x7f6d581c8fb7]", > "(abort()+0x141) [0x7f6d581ca921]", > "(ceph::__ceph_abort(char const*, int, char const*, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&)+0x1b2) [0x55ddc61f245f]", > "(BlueStore::ExtentMap::allocate_spanning_blob_id()+0x104) [0x55ddc674b594]", > "(BlueStore::ExtentMap::reshard(KeyValueDB*, std::shared_ptr<KeyValueDB::TransactionImpl>)+0x1408) [0x55ddc674c9c8]", > "(BlueStore::_record_onode(boost::intrusive_ptr<BlueStore::Onode>&, std::shared_ptr<KeyValueDB::TransactionImpl>&)+0x91c) [0x55ddc674f4ec]", > "(BlueStore::_txc_write_nodes(BlueStore::TransContext*, std::shared_ptr<KeyValueDB::TransactionImpl>)+0x7e) [0x55ddc6751b4e]", > "(BlueStore::queue_transactions(boost::intrusive_ptr<ObjectStore::CollectionImpl>&, std::vector<ceph::os::Transaction, std::allocator<ceph::os::Transaction> >&, boost::intrusive_ptr<TrackedOp>, ThreadPool::TPHandle*)+0x2fc) [0x55ddc677892c]", > "(non-virtual thunk to PrimaryLogPG::queue_transactions(std::vector<ceph::os::Transaction, std::allocator<ceph::os::Transaction> >&, boost::intrusive_ptr<OpRequest>)+0x54) [0x55ddc63eef44]", > "(ECBackend::handle_sub_write(pg_shard_t, boost::intrusive_ptr<OpRequest>, ECSubWrite&, ZTracer::Trace const&)+0x9cd) [0x55ddc65cb95d]", > "(ECBackend::_handle_message(boost::intrusive_ptr<OpRequest>)+0x23d) [0x55ddc65e3c2d]", > "(PGBackend::handle_message(boost::intrusive_ptr<OpRequest>)+0x97) [0x55ddc643b157]", > "(PrimaryLogPG::do_request(boost::intrusive_ptr<OpRequest>&, ThreadPool::TPHandle&)+0x6fd) [0x55ddc63ddddd]", > "(OSD::dequeue_op(boost::intrusive_ptr<PG>, boost::intrusive_ptr<OpRequest>, ThreadPool::TPHandle&)+0x17b) [0x55ddc62618bb]", > "(ceph::osd::scheduler::PGOpItem::run(OSD*, OSDShard*, boost::intrusive_ptr<PG>&, ThreadPool::TPHandle&)+0x67) [0x55ddc64bf167]", > "(OSD::ShardedOpWQ::_process(unsigned int, ceph::heartbeat_handle_d*)+0x90c) [0x55ddc627ef4c]", > "(ShardedThreadPool::shardedthreadpool_worker(unsigned int)+0x4ac) [0x55ddc68d1d0c]", > "(ShardedThreadPool::WorkThreadSharded::entry()+0x10) [0x55ddc68d4f60]", > "(()+0x76db) [0x7f6d5950b6db]", > "(clone()+0x3f) [0x7f6d582ab71f]" > ] > > > Best Regards > > > > > > > > _______________________________________________ > ceph-users mailing list -- ceph-users@xxxxxxx<mailto:ceph-users@xxxxxxx> > To unsubscribe send an email to ceph-users-leave@xxxxxxx<mailto:ceph-users-leave@xxxxxxx> -- Igor Fedotov Ceph Lead Developer _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |