[Ceph-announce] v10.2.3 Jewel Released

This point release fixes several important bugs in RBD mirroring, RGW multi-site, CephFS, and RADOS. We recommend that all v10.2.x users upgrade. Notable changes in this release include: * build/ops: 60-ceph-partuuid-workaround-rules still needed by debian jessie (udev 215-17) (#16351, runsisi, Loic Dachary) * build/ops: ceph Resource Agent does not work with systemd (#14828, Nathan Cutler) * build/ops: ceph-base requires parted (#16095, Ken Dreyer) * build/ops: ceph-osd-prestart.sh contains Upstart-specific code (#15984, Nathan Cutler) * build/ops: mount.ceph: move from ceph-base to ceph-common and add symlink in /sbin for SUSE (#16598, #16645, Nathan Cutler, Dan Horák, Ricardo Dias, Kefu Chai) * build/ops: need rocksdb commit 7ca731b12ce for ppc64le build (#17092, Nathan Cutler) * build/ops: rpm: OBS needs ExclusiveArch (#16936, Michel Normand) * cli: ceph command line tool chokes on ceph –w (the dash is unicode 'en dash' &ndash, copy-paste to reproduce) (#12287, Oleh Prypin, Kefu Chai) * common: expose buffer const_iterator symbols (#16899, Noah Watkins) * common: global-init: fixup chown of the run directory along with log and asok files (#15607, Karol Mroz) * fs: ceph-fuse: link to libtcmalloc or jemalloc (#16655, Yan, Zheng) * fs: client: crash in unmount when fuse_use_invalidate_cb is enabled (#16137, Yan, Zheng) * fs: client: fstat cap release (#15723, Yan, Zheng, Noah Watkins) * fs: essential backports for OpenStack Manila (#15406, #15614, #15615, John Spray, Ramana Raja, Xiaoxi Chen) * fs: fix double-unlock on shutdown (#17126, Greg Farnum) * fs: fix mdsmap print_summary with standby replays (#15705, John Spray) * fs: fuse mounted file systems fails SAMBA CTDB ping_pong rw test with v9.0.2 (#12653, #15634, Yan, Zheng) * librados: Add cleanup message with time to rados bench output (#15704, Vikhyat Umrao) * librados: Missing export for rados_aio_get_version in src/include/rados/librados.h (#15535, Jim Wright) * librados: osd: bad flags can crash the osd (#16012, Sage Weil) * librbd: Close journal and object map before flagging exclusive lock as released (#16450, Jason Dillaman) * librbd: Crash when utilizing advisory locking API functions (#16364, Jason Dillaman) * librbd: ExclusiveLock object leaked when switching to snapshot (#16446, Jason Dillaman) * librbd: FAILED assert(object_no < m_object_map.size()) (#16561, Jason Dillaman) * librbd: Image removal doesn't necessarily clean up all rbd_mirroring entries (#16471, Jason Dillaman) * librbd: Object map/fast-diff invalidated if journal replays the same snap remove event (#16350, Jason Dillaman) * librbd: Timeout sending mirroring notification shouldn't result in failure (#16470, Jason Dillaman) * librbd: Whitelist EBUSY error from snap unprotect for journal replay (#16445, Jason Dillaman) * librbd: cancel all tasks should wait until finisher is done (#16517, Haomai Wang) * librbd: delay acquiring lock if image watch has failed (#16923, Jason Dillaman) * librbd: fix missing return statement if failed to get mirror image state (#16600, runsisi) * librbd: flag image as updated after proxying maintenance op (#16404, Jason Dillaman) * librbd: mkfs.xfs slow performance with discards and object map (#16707, #16689, Jason Dillaman) * librbd: potential use after free on refresh error (#16519, Mykola Golub) * librbd: rbd-nbd does not properly handle resize notifications (#15715, Mykola Golub) * librbd: the option 'rbd_cache_writethrough_until_flush=true' dosn't work (#16740, #16386, #16708, #16654, #16478, Mykola Golub, xinxin shu, Xiaowei Chen, Jason Dillaman) * mds: tell command blocks forever with async messenger (TestVolumeClient.test_evict_client failure) (#16288, Douglas Fuller) * mds: Confusing MDS log message when shut down with stalled journaler reads (#15689, John Spray) * mds: Deadlock on shutdown active rank while busy with metadata IO (#16042, Patrick Donnelly) * mds: Failing file operations on kernel based cephfs mount point leaves unaccessible file behind on hammer 0.94.7 (#16013, Yan, Zheng) * mds: Fix shutting down mds timed-out due to deadlock (#16396, Zhi Zhang) * mds: MDSMonitor fixes (#16136, xie xingguo) * mds: MDSMonitor::check_subs() is very buggy (#16022, Yan, Zheng) * mds: Session::check_access() is buggy (#16358, Yan, Zheng) * mds: StrayManager.cc: 520: FAILED assert(dnl->is_primary()) (#15920, Yan, Zheng) * mds: enforce a dirfrag limit on entries (#16164, Patrick Donnelly) * mds: fix SnapRealm::have_past_parents_open() (#16299, Yan, Zheng) * mds: fix getattr starve setattr (#16154, Yan, Zheng) * mds: wrongly treat symlink inode as normal file/dir when symlink inode is stale on kcephfs (#15702, Zhi Zhang) * mon: "mon metadata" fails when only one monitor exists (#15866, John Spray, Kefu Chai) * mon: Monitor: validate prefix on handle_command() (#16297, You Ji) * mon: OSDMonitor: drop pg temps from not the current primary (#16127, Samuel Just) * mon: prepare_pgtemp needs to only update up_thru if newer than the existing one (#16185, Samuel Just) * msgr: AsyncConnection::lockmsg/async lockdep cycle: AsyncMessenger::lock, MDSDaemon::mds_lock, AsyncConnection::lock (#16237, Haomai Wang) * msgr: async messenger mon crash (#16378, #16418, Haomai Wang) * msgr: backports of all asyncmsgr fixes to jewel (#15503, #15372, Yan Jun, Haomai Wang, Piotr Dałek) * msgr: msg/async: connection race hang (#15849, Haomai Wang) * osd: FileStore: umount hang because sync thread doesn't exit (#15695, Kefu Chai) * osd: Fixes for list-inconsistent-* (#15766, #16192, #15719, David Zafman) * osd: New pools have bogus stuck inactive/unclean HEALTH_ERR messages until they are first active and clean (#14952, Sage Weil) * osd: OSD crash with Hammer to Jewel Upgrade: void FileStore::init_temp_collections() (#16672, David Zafman) * osd: OSD failed to subscribe skipped osdmaps after ceph osd pause (#17023, Kefu Chai) * osd: ObjectCacher split BufferHead read fix (#16002, Greg Farnum) * osd: ReplicatedBackend doesn't increment stats on pull, only push (#16277, Kefu Chai) * osd: Scrub error: 0/1 pinned (#15952, Samuel Just) * osd: crash adding snap to purged_snaps in ReplicatedPG::WaitingOnReplicas (#15943, Samuel Just) * osd: partprobe intermittent issues during ceph-disk prepare (#15176, Marius Vollmer, Loic Dachary) * osd: saw valgrind issues in ReplicatedPG::new_repop (#16801, Kefu Chai) * osd: sparse_read on ec pool should return extends with correct offset (#16138, kofiliu) * osd:sched_time not actually randomized (#15890, xie xingguo) * rbd: ImageReplayer::is_replaying does not include flush state (#16970, Jason Dillaman) * rbd: Journal duplicate op detection can cause lockdep error (#16363, Jason Dillaman) * rbd: Journal needs to handle duplicate maintenance op tids (#16362, Jason Dillaman) * rbd: Unable to disable journaling feature if in unexpected mirror state (#16348, Jason Dillaman) * rbd: bashism in src/rbdmap (#16608, Jason Dillaman) * rbd: doc: format 2 now is the default image format (#17026, Chengwei Yang) * rbd: hen journaling is enabled, a flush request shouldn't flush the cache (#15761, Yuan Zhou) * rbd: possible race condition during journal transition from replay to ready (#16198, Jason Dillaman) * rbd: qa/workunits/rbd: respect RBD_CREATE_ARGS environment variable (#16289, Mykola Golub) * rbd: rbd-mirror should disable proxied maintenance ops for non-primary image (#16411, Jason Dillaman) * rbd: rbd-mirror: FAILED assert(m_local_image_ctx->object_map != nullptr) (#16558, Jason Dillaman) * rbd: rbd-mirror: FAILED assert(m_on_update_status_finish == nullptr) (#16956, Jason Dillaman) * rbd: rbd-mirror: FAILED assert(m_state == STATE_STOPPING) (#16980, Jason Dillaman) * rbd: rbd-mirror: ensure replay status formatter has completed before stopping replay (#16352, Jason Dillaman) * rbd: rbd-mirror: include local pool id in resync throttle unique key (#16536, #15239, #16488, #16491, #16329, #15108, #15670, Ricardo Dias, Jason Dillaman) * rbd: rbd-mirror: potential race condition accessing local image journal (#16230, Jason Dillaman) * rbd: rbd-mirror: reduce memory footprint during journal replay (#16321, #16489, #16622, #16539, #16223, #16349, Mykola Golub, Jason Dillaman) * rgw: A query on a static large object fails with 404 error (#16015, Radoslaw Zarzynski) * rgw: Add zone rename to radosgw_admin (#16934, Shilpa Jagannath) * rgw: Bucket index shards orphaned after bucket delete (#16412, Orit Wasserman) * rgw: Bug when using port 443s in rgw. (#16548, Pritha Srivastava) * rgw: Fallback to Host header for bucket name. (#15975, Robin H. Johnson) * rgw: Fix civetweb IPv6 (#16928, Robin H. Johnson) * rgw: Increase log level for messages occuring while running rgw admin command (#16935, Shilpa Jagannath) * rgw: No Last-Modified, Content-Size and X-Object-Manifest headers if no segments in DLO manifest (#15812, Radoslaw Zarzynski) * rgw: RGWPeriodPuller tries to pull from itself (#16939, Casey Bodley) * rgw: Set Access-Control-Allow-Origin to a Asterisk if allowed in a rule (#15348, Wido den Hollander) * rgw: Swift API returns double space usage and objects of account metadata (#16188, Albert Tu) * rgw: account/container metadata not actually present in a request are deleted during POST through Swift API (#15977, #15779, Radoslaw Zarzynski) * rgw: add socket backlog setting for via ceph.conf (#16406, Feng Guo) * rgw: add tenant support to multisite sync (#16469, #16121, #16665, Yehuda Sadeh, Josh Durgin, Casey Bodley, Pritha Srivastava) * rgw: add_zone only clears master_zone if --master=false (#15901, Casey Bodley) * rgw: aws4 parsing issue (#15940, #15939, Yehuda Sadeh) * rgw: aws4: add STREAMING-AWS4-HMAC-SHA256-PAYLOAD support (#16146, Radoslaw Zarzynski, Javier M. Mellid) * rgw: backport merge of static sites fixes (#15555, #15532, #15531, Robin H. Johnson) * rgw: can set negative max_buckets on RGWUserInfo (#14534, Yehuda Sadeh) * rgw: cleanup radosgw-admin temp command as it was deprecated (#16023, Vikhyat Umrao) * rgw: comparing return code to ERR_NOT_MODIFIED in rgw_rest_s3.cc (needs minus sign) (#16327, Nathan Cutler) * rgw: custom metadata aren't camelcased in Swift's responses (#15902, Radoslaw Zarzynski) * rgw: data sync stops after getting error in all data log sync shards (#16530, Yehuda Sadeh) * rgw: default zone and zonegroup cannot be added to a realm (#16839, Casey Bodley) * rgw: document multi tenancy (#16635, Pete Zaitcev) * rgw: don't unregister request if request is not connected to manager (#15911, Yehuda Sadeh) * rgw: failed to create bucket after upgrade from hammer to jewel (#16627, Orit Wasserman) * rgw: fix ldap bindpw parsing (#16286, Matt Benjamin) * rgw: fix multi-delete query param parsing. (#16618, Robin H. Johnson) * rgw: improve support for Swift's object versioning. (#15925, Radoslaw Zarzynski) * rgw: initial slashes are not properly handled in Swift's BulkDelete (#15948, Radoslaw Zarzynski) * rgw: master: build failures with boost > 1.58 (#16392, #16391, Abhishek Lekshmanan) * rgw: multisite segfault on ~RGWRealmWatcher if realm was deleted (#16817, Casey Bodley) * rgw: multisite sync races with deletes (#16222, #16464, #16220, #16143, Yehuda Sadeh, Casey Bodley) * rgw: multisite: preserve zone's extra pool (#16712, Abhishek Lekshmanan) * rgw: object expirer's hints might be trimmed without processing in some circumstances (#16705, #16684, Radoslaw Zarzynski) * rgw: radosgw-admin failure for user create after upgrade from hammer to jewel (#15937, Orit Wasserman, Abhishek Lekshmanan) * rgw: radosgw-admin: EEXIST messages for create operations (#15720, Abhishek Lekshmanan) * rgw: radosgw-admin: inconsistency in uid/email handling (#13598, Matt Benjamin) * rgw: realm pull fails when using apache frontend (#15846, Orit Wasserman) * rgw: retry on bucket sync errors (#16108, Yehuda Sadeh) * rgw: s3website: x-amz-website-redirect-location header returns malformed HTTP response (#15531, Robin H. Johnson) * rgw: segfault in RGWOp_MDLog_Notify (#16666, Casey Bodley) * rgw: segmentation fault on error_repo in data sync (#16603, Casey Bodley) * rgw: selinux denials in RGW (#16126, Boris Ranto) * rgw: support size suffixes for --max-size in radosgw-admin command (#16004, Vikhyat Umrao) * rgw: updating CORS/ACLs might not work in some circumstances (#15976, Radoslaw Zarzynski) * rgw: use zone endpoints instead of zonegroup endpoints (#16834, Casey Bodley) * tests: improve rbd-mirror test case coverage (#16197, Mykola Golub, Jason Dillaman) * tests: rados/test.sh workunit timesout on OpenStack (#15403, Loic Dachary) * tools: ceph-disk: Accept bcache devices as data disks (#13278, Peter Sabaini) * tools: src/script/subman fails with KeyError: 'nband' (#16961, Loic Dachary, Ali Maredia) For more detailed information refer to the complete changelog[1] and the release notes[2] Getting Ceph ------------ * Git at git://github.com/ceph/ceph.git * Tarball at http://download.ceph.com/tarballs/ceph-10.2.3.tar.gz * For packages, see http://ceph.com/docs/master/install/get-packages * For ceph-deploy, see http://ceph.com/docs/master/install/install-ceph-deploy [1]: http://docs.ceph.com/docs/master/_downloads/v10.2.3.txt [2]: http://docs.ceph.com/docs/master/release-notes/#v10-2-3-jewel Regards, Abhishek -- SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)