We're happy to announce the third bugfix release of Luminous v12.2.x
long term stable release series. It contains a range of bug fixes and
stability improvements across Bluestore, CephFS, RBD & RGW. We recommend all the
users of 12.2.x series update.
Notable Changes
---
*CephFS*:
* The CephFS client now checks for older kernels' inability to reliably clear
dentries from the kernel dentry cache. The new option
client_die_on_failed_dentry_invalidate (default: true) may be turned off to
allow the client to proceed (dangerous!).
For the full changelog with links, please refer to the release blog at
http://ceph.com/releases/v12-2-3-luminous-released/. We thank everyone
for contributing PRs, reporting bugs and helping make luminous
luminescent. Many thanks to Yuri for keeping the QE runs across the
components green.
Getting Ceph
* Git at git://github.com/ceph/ceph.git
* Tarball at http://download.ceph.com/tarballs/ceph-12.2.3.tar.gz
* For packages, see http://docs.ceph.com/docs/master/install/get-packages/
* For ceph-deploy, see
http://docs.ceph.com/docs/master/install/install-ceph-deploy
* Release git sha1: 2dab17a455c09584f2a85e6b10888337d1ec8949
Changelog
-
* bluestore: do not crash on over-large objects (issue#22161, pr#19630, Sage
Weil)
* bluestore: OSD crash on boot with assert caused by Bluefs on flush write
(issue#21932, pr#19047, Jianpeng Ma)
* build/ops: ceph-base symbols not stripped in debs (issue#22640, pr#19969,
Sage Weil)
* build/ops: ceph-conf: dump parsed config in plain text or as json
(issue#21862, pr#18842, Piotr Dałek)
* build/ops: ceph-mgr dashboard has dependency on python-jinja2 (issue#22457,
pr#19865, John Spray)
* build/ops: ceph-volume fails when centos7 image doesn't have lvm2 installed
(issue#22443, issue#22217, pr#20215, Nathan Cutler, Theofilos Mouratidis)
* build/ops: Default kernel.pid_max is easily exceeded during recovery on high
OSD-count system (issue#21929, pr#19133, David Disseldorp, Kefu Chai)
* build/ops: install-deps.sh: revert gcc to the one shipped by distro
(issue#0, pr#19680, Kefu Chai)
* build/ops: luminous build fails with --without-radosgw (issue#22321,
pr#19483, Jason Dillaman)
* build/ops: move ceph-\*-tool binaries out of ceph-test subpackage
(issue#22319, issue#21762, pr#19355, liuchang0812, Nathan Cutler, Kefu Chai,
Sage Weil)
* build.ops: rpm: adjust ceph-{osdomap,kvstore,monstore}-tool feature move
(issue#22558, pr#19839, Kefu Chai)
* ceph: cluster [ERR] Unhandled exception from module 'balancer' while running
on mgr.x: 'NoneType' object has no attribute 'iteritems'" in cluster log
(issue#22090, pr#19023, Sage Weil)
* cephfs: cephfs-journal-tool: add "set pool_id" option (issue#22631, pr#20085,
dongdong tao)
* cephfs: cephfs-journal-tool: tool would miss to report some invalid range
(issue#22459, pr#19626, dongdong tao)
* cephfs: cephfs: potential adjust failure in lru_expire (issue#22458,
pr#19627, dongdong tao)
* cephfs: "ceph tell mds" commands result in "File exists" errors on client
admin socket (issue#21406, issue#21967, pr#18831, Patrick Donnelly)
* cephfs: client: anchor Inode while trimming caps (issue#22157, pr#19105,
Patrick Donnelly)
* cephfs: client: avoid recursive lock in ll_get_vino (issue#22629, pr#20086,
dongdong tao)
* cephfs: client: dual client segfault with racing ceph_shutdown (issue#21512,
issue#20988, pr#20082, Jeff Layton)
* cephfs: client: implement delegation support in userland cephfs (issue#18490,
pr#19480, Jeff Layton)
* cephfs: client: quit on failed remount during dentry invalidate test #19218
(issue#22269, pr#19370, Patrick Donnelly)
* cephfs: List of filesystems does not get refreshed after a filesystem
deletion (issue#21599, pr#18730, John Spray)
* cephfs: MDS : Avoid the assert failure when the inode for the cap_export from
other… (issue#22610, pr#20300, Jianyu Li)
* cephfs: MDSMonitor: monitor gives constant "is now active in filesystem
cephfs as rank" cluster log info messages (issue#21959, pr#19055, Patrick
Donnelly)
* cephfs: racy is_mounted() checks in libcephfs (issue#21025, pr#17875, Jeff
Layton)
* cephfs: src/mds/MDCache.cc: 7421: FAILED assert(CInode::count() ==
inode_map.size() + snap_inode_map.size()) (issue#21928, pr#18912, "Yan, Zheng")
* cephfs: vstart_runner: fixes for recent cephfs changes (issue#22526,
pr#19829, Patrick Donnelly)
* ceph-volume: adds a --destroy flag to ceph-volume lvm zap (issue#22653,
pr#20240, Andrew Schoen)
* ceph-volume: adds success messages for lvm prepare/activate/create
(issue#22307, pr#20238, Andrew Schoen)
* ceph-volume: dmcrypt support for lvm (issue#22619, pr#20241, Alfredo Deza)
* ceph-volume dmcrypt support for simple (issue#22620, pr#20350, Andrew Schoen,
Alfredo Deza)
* ceph-volume: do not use --key during mkfs (issue#22283, pr#20244, Kefu Chai,
Sage Weil)
* ceph-volume: fix usage of the --osd-id flag (issue#22642, issue#22836,
pr#20323, Andrew Schoen)
* ceph-volume Format