Merge pull request #16043 from jcsp/wip-dashboard-updates

mgr: dashboard improvements

Reviewed-by: Kefu Chai <kchai@redhat.com>

mgr/dashboard: update health display

This takes account of the new health format, also
expands and visually cleans up the frontpage
where we put the health information.

Dark backgrounds make it much easier to use
red/amber/green colours to grab attention.

Signed-off-by: John Spray <john.spray@redhat.com>

mgr: expose a MgrMap in PyModules

Signed-off-by: John Spray <john.spray@redhat.com>

Merge pull request #16020 from jcsp/wip-20383

mgr: clean up daemon start process

Reviewed-by: Sage Weil <sage@redhat.com>

Merge pull request #16338 from scienceluo/wip-doc-branch

doc/release-notes: Luminous release notes typo fixes

Reviewed-by: Jos Collin <jcollin@redhat.com>

doc/release-notes: Luminous release notes typo fixes "systemctl ceph-osd.target"->"systemctl restart ceph-osd.target" and "systemctl ceph-mgr.target"->"systemctl restart ceph-mgr.target"

Signed-off-by: Luo Kexue <luo.kexue@zte.com.cn>

Merge pull request #16318 from smithfarm/wip-jewel-10-2-9

doc: Jewel v10.2.9 release notes

Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>

Merge pull request #16330 from scienceluo/master

doc/release-notes: Luminous release notes typo fixes  "ceph config-key ls"->"ceph config-key list"

Reviewed-by: xie xingguo <xie.xingguo@zte.com.cn>
Reviewed-by: Kefu Chai <kchai@redhat.com>

doc:fix release-notes

Signed-off-by: luo.kexue@zte.com.cn

Merge pull request #16300 from liewegas/wip-20600

mon: fix hang on deprecated/removed 'pg set_*full_ratio' commands

Reviewed-by: Greg Farnum <gfarnum@redhat.com>

Merge pull request #16321 from Yan-waller/wip-walle-0712cephosd

common: misc cleanups in common, global, os, osd submodules

Reviewed-by: Jos Collin <jcollin@redhat.com>

Merge pull request #16322 from liewegas/wip-20617

qa/tasks/ceph_manager: wait longer for pg stats to flush

Reviewed-by: Kefu Chai <kchai@redhat.com>

Merge pull request #16323 from ceph/revert-15897-wip-20390

Revert "msg/async: increase worker reference with local listen table enabled backend"

Merge pull request #16319 from tchaikov/wip-ceph-helper-with-exp-features

qa/workunits/ceph-helpers: enable experimental features for osd

Reviewed-by: Loic Dachary <ldachary@redhat.com>

Merge pull request #16320 from tchaikov/wip-clang-analyzer-warnings

test,mon,msg: kill clang analyzer warnings

Reviewed-by: Haomai Wang <haomai@xsky.com>
Reviewed-by: Brad Hubbard <bhubbard@redhat.com>

doc: Jewel v10.2.9 changelog

Signed-off-by: Nathan Cutler <ncutler@suse.com>

mon/PGMonitor: EOPNOTSUPP for old pgmon commands

This includes 'pg set_full_ratio', which we have only for the upgrade, but
goes away afterwards.

Also, return true to either swallow the request or indicate it has been
processed.

Fixes: http://tracker.ceph.com/issues/20600
Signed-off-by: Sage Weil <sage@redhat.com>

mon: do not assign to never-read variable

this silences clang analyzer's warning of

Value stored to 'err' is never read

Signed-off-by: Kefu Chai <kchai@redhat.com>

msg/async/rdma: return stored errno on error

otherwise the errno would be overwritten, and we are returning 0 or the
errno set by ::close()

Signed-off-by: Kefu Chai <kchai@redhat.com>

test: test_c_headers: silence clang analyzer warnings

this silences clang analyzer's warnings like:

Value stored to 'ret' is never read

Signed-off-by: Kefu Chai <kchai@redhat.com>

Merge pull request #16262 from liewegas/wip-20208

mgr/ClusterState: do not mangle PGMap outside of Incremental

Merge pull request #9974 from weiqiaomiao/wqm-wip-copy_obj

rgw: fix memory leak in copy_obj_to_remote_dest

Reviewed-by: Orit Wasserman <owasserm@redhat.com>

Merge pull request #11124 from zhangsw/cleanup-rgwrados-deleteobj

rgw: remove a redundant judgement in rgw_rados.cc:delete_obj.

Reviewed-by: Yehuda Sadeh <yehuda@redhat.com>

Merge pull request #12010 from zhangsw/fix-rgw-multipart-bug

rgw: Fix a bug that multipart upload may exceed the quota.

Reviewed-by: Yehuda Sadeh <yehuda@redhat.com>

Merge pull request #12197 from zhangsw/fix-rgw-metasync-lock-bug

rgw: lock is not released when set sync marker is failed.

Reviewed-by: Casey Bodley <cbodley@redhat.com>

qa/tasks/ceph_manager: wait longer for pg stats to flush

An ill-timed mgr restart could blow the current 15s wait.

Signed-off-by: Sage Weil <sage@redhat.com>

Merge pull request #16021 from joscollin/wip-uninitialized-pointer-fields-1

rgw: Initialize pointer fields

Reviewed-by: Yehuda Sadeh <yehuda@redhat.com>

Merge pull request #16014 from xiexingguo/wip-autoclass

osd/OSD: auto class on osd start up

Reviewed-by: Sage Weil <sage@redhat.com>

Merge pull request #15774 from shashalu/drop-temp-var

rgw: using RGW_OBJ_NS_MULTIPART in check_bad_index_multipart

Reviewed-by: Casey Bodley <cbodley@redhat.com>

doc: Jewel v10.2.9 release notes

Signed-off-by: Nathan Cutler <ncutler@suse.com>

Revert "msg/async: increase worker reference with local listen table enabled backend"

Merge pull request #16255 from trociny/wip-test-librbd-internals

test/librbd: re-enable internal tests in ceph_test_librbd

Reviewed-by: Jason Dillaman <dillaman@redhat.com>

Merge pull request #16130 from fullerdj/wip-djf-ceph-connect-timeout

ceph.in: Check return value when connecting

Reviewed-by: Kefu Chai <kchai@redhat.com>
Reviewed-by: Alfredo Deza <adeza@redhat.com>

Merge pull request #16243 from markhpc/wip-bluestore-freelist-iterator

os/bluestore: Make BitmapFreelistManager kv itereator short lived.

Reviewed-by: Sage Weil <sage@redhat.com>

Merge pull request #16269 from liewegas/wip-bluestore-deferred-pending

os/bluestore: only submit deferred if there is any

Reviewed-by: xie xingguo <xie.xingguo@zte.com.cn>
Reviewed-by: Mark Nelson <mnelson@redhat.com>

Merge pull request #16306 from liewegas/wip-reg11184-health

qa/suites/rados/singleton/all/reg11184: whitelist health warnings

osd: cleanups

Signed-off-by: Yan Jun <yan.jun8@zte.com.cn>

Merge pull request #16317 from tchaikov/wip-0-osd-is-not-an-error

qa/workunits/ceph-helpers: test wait_for_health_ok differently

Reviewed-by: Loic Dachary <ldachary@redhat.com>

Merge pull request #16274 from smithfarm/wip-jewel-10-2-8

doc: Jewel v10.2.8 release notes

Reviewed-by: Sage Weil <sage@redhat.com>
Reviewed-by: Gregory Farnum <gfarnum@redhat.com>

Merge pull request #16267 from liewegas/wip-restful-defaults

mgr/restful: bind to :: and update docs

Reviewed-by: John Spray <john.spray@redhat.com>
Reviewed-by: Boris Ranto <branto@redhat.com>

Merge pull request #16311 from Songweibin/wip-update-info-affi

.mailmap, .organizationmap: Update Song Weibin information and affiliation

Reviewed-by: Abhishek Lekshmanan <abhishek@suse.com>

Merge pull request #16303 from bstillwell/releases-doc-update-201707

doc/releases: Update releases from Feb 2017 to July 2017

Reviewed-by: Kefu Chai <kchai@redhat.com>

doc: add v10.2.8 changelog

Signed-off-by: Nathan Cutler <ncutler@suse.com>

qa/workunits/ceph-helpers: enable experimental features for osd

it matches the settings in vstart.sh, also it would be handy for those
who are still developing on btrfs, which is now marked as an experimental
features now.

Signed-off-by: Kefu Chai <kchai@redhat.com>

qa/workunits/ceph-helpers: test wait_for_health_ok differently

0 OSDs is not an error anymore in the new health checking implemented by
OSDMap::check_health(). this case was treated as an error before, see
OSDMonitor::get_health(). but an osdmap without any OSD is fine, i
think. but an osdmap with 3 OSDs, but all of them are down and out, this
is an error. and we do report this as an error. so, let's update the
test instead.

Signed-off-by: Kefu Chai <kchai@redhat.com>

Merge pull request #16277 from smithfarm/wip-rn-script

tools: ceph-release-notes: handle an edge case

Reviewed-by: Kefu Chai <kchai@redhat.com>

doc: release notes: note MDS regression in 10.2.8

See the discussion in https://github.com/ceph/ceph/pull/16192

Signed-off-by: Nathan Cutler <ncutler@suse.com>

Merge pull request #16264 from dillaman/wip-20571

rbd-mirror: ignore permission errors on rbd_mirroring object

Reviewed-by: Mykola Golub <mgolub@mirantis.com>

.mailmap, .organizationmap: Update Song Weibin information and affiliation

Signed-off-by: songweibin <song.weibin@zte.com.cn>

qa/suites/rados/singleton/all/reg11184: whitelist health warnings

Signed-off-by: Sage Weil <sage@redhat.com>

Update releases from Feb 2017 to July 2017

Pull in the latest releases from the past 5 months and fix some of the
links so they jump to the correct sections in the release notes.

Signed-off-by: Bryan Stillwell <bstillwell@godaddy.com>

Merge pull request #16301 from dmick/master

mgr: increase debug level for ticks 0 -> 10

Mgr: increase debug level for ticks 0 -> 10

Signed-off-by: Dan Mick <dmick@redhat.com>

mon/MonCommands: mark 'pg set_*_ratio' deprecated

Signed-off-by: Sage Weil <sage@redhat.com>

Merge pull request #16263 from liupan1111/wip-fix-fio

test/fio: remove experimental option for bluestore & rocksdb.

Reviewed-by: Casey Bodley <cbodley@redhat.com>

Merge pull request #15643 from liewegas/wip-health

mon: revamp health check/warning system

mon/PGMap: adjust scrub checks to avoid overflow for future stamps

Avoid an overflow (and false warning) when scrub stamps are in the future.

Signed-off-by: Sage Weil <sage@redhat.com>

qa/workunits/cephtool/test.sh: adjust full tests to avoid races

OSDs may report fullness in any order.

Signed-off-by: Sage Weil <sage@redhat.com>

qa/tasks/ceph: wait for osds to come up before creating pool

Avoid health warnings.

Signed-off-by: Sage Weil <sage@redhat.com>

qa/tasks/ceph_test_case.py: update health check helpers

Signed-off-by: Sage Weil <sage@redhat.com>

qa/suites/fs: whitelist health warnings

Signed-off-by: Sage Weil <sage@redhat.com>

qa/suites/rgw/thrash: whitelist

Signed-off-by: Sage Weil <sage@redhat.com>

qa/suites/rbd: whitelist health messages

Signed-off-by: Sage Weil <sage@redhat.com>

qa: whitelist health warnings

Signed-off-by: Sage Weil <sage@redhat.com>

qa/workunits/cephtool/test.sh: adjust for new health error codes

Signed-off-by: Sage Weil <sage@redhat.com>

mon/MgrMonitor: clear last_beacon after mon election

The last_beacon map is local to an election interval; if there is a new
election completed we should reset it or else we may kill an apparently
laggy mgr that hasn't been able to get a beacon processed due to the mon
quorum changing, or had its beacon processed on a different leader.

Signed-off-by: Sage Weil <sage@redhat.com>

mon: clean up `osd out` messages

Cleaner prose for the auto-out case, and add
a cluster log message for OSDs that go out
at the behest of the administrator.

Signed-off-by: John Spray <john.spray@redhat.com>

osd: don't log per-PG backfill messages at INFO level

This behaviour led to way too many messages going to
the cluster log when an OSD is marked in.  Retain
the messages at debug level.

Signed-off-by: John Spray <john.spray@redhat.com>

mon: simplify PG health checks

Instead of a distinct health check for each possible
PG state, group the states into categories for availability,
degraded, damage, and report on that.

That way, while a PG/pool is suffering from one of those
bad PG states, health conditions don't keep toggling on and
off as we transition from one unavailable state to another
unavailable state.

Signed-off-by: John Spray <john.spray@redhat.com>

mgr/dashboard: update for new style health checks

Signed-off-by: John Spray <john.spray@redhat.com>

mon: demote cluster map prints to DEBUG level

The PaxosService subclasses should be writing out
informative log messages, and not relying on
a stream of map summary prints to communicate
changes.

Signed-off-by: John Spray <john.spray@redhat.com>

mon: prettify health check log messages

Add a "Cluster is now healthy" to give clarity
after a series of "health check cleared" that
they were the last ones.

Convert certain health check messages into
well formed sentences.

Don't print severity in the log string (it's already
expressed in the severity of the log entry.

Signed-off-by: John Spray <john.spray@redhat.com>

mgr: fix spurious PG health messages on mgr restart

Previously, the mgr would send MMonMgrReport indicating
a very unhappy PGMap to the mon right after startup.

This is a change to hold off on sending that report until
all the OSDs have reported in, or until some time has passed.

Signed-off-by: John Spray <john.spray@redhat.com>

mon: don't consider a starting mgr to be an error

The .available flag is there to tell MgrClients whether
to try and connect -- it isn't the right condition
for a health complaint.

Signed-off-by: John Spray <john.spray@redhat.com>

mon: pass new style health to mgr

Signed-off-by: John Spray <john.spray@redhat.com>

mon: prefix periodic health reminder with 'overall'

...so we can whitelist it.

Signed-off-by: Sage Weil <sage@redhat.com>

mon/PGMap: rename a few health checks

Signed-off-by: Sage Weil <sage@redhat.com>

mgr/DaemonServer: debug log health checks

Signed-off-by: Sage Weil <sage@redhat.com>

mon/MgrStatMonitor: show health check count on receipt

Signed-off-by: Sage Weil <sage@redhat.com>

messages/MMonMgrReport: show health check count

Signed-off-by: Sage Weil <sage@redhat.com>

mon/PGMap: only warn about too few pgs after >0 pools exist

Signed-off-by: Sage Weil <sage@redhat.com>

mon/PGMap: some stuck warnings are err, some warn

inactive and stale -> error
degraded, unclean, undersized -> warning

Signed-off-by: Sage Weil <sage@redhat.com>

qa/suites/rados: whitelist health warnings

Signed-off-by: Sage Weil <sage@redhat.com>

mon/PGMap: do not warn about recovering, peering, stale

Wait for stuck before complaining.  These aren't scary in and of
themselves.

Signed-off-by: Sage Weil <sage@redhat.com>

qa/tasks/mon_clock_skew_check: vastly simplify

Signed-off-by: Sage Weil <sage@redhat.com>

mon: shorten 'stuck' threshold from 5m -> 1m

Signed-off-by: Sage Weil <sage@redhat.com>

osd/OSDMap: add per-osd flag OSD_FLAGS check

rename OSD_FLAGS to OSDMAP_FLAGS

Signed-off-by: Sage Weil <sage@redhat.com>

osd/OSDMap: rename a few health checks

Signed-off-by: Sage Weil <sage@redhat.com>

mon: move osd health checks into OSDMap method

...with one check moving into HealthMonitor where it
belongs.

Signed-off-by: Sage Weil <sage@redhat.com>

qa/tasks/ceph: stop logging health on shutdown

Don't log health during actual teardown or we'll see
various scary messages unrelated to our test run.

Signed-off-by: Sage Weil <sage@redhat.com>

mon/MgrMonitor: do not issue MGR_DOWN on new cluster

It is normal for the initial cluster to lack a mgr.  Wait for some
grace period before complaining about a missing mgr.

Default to 30m.

Signed-off-by: Sage Weil <sage@redhat.com>

mon/Monitor: periodically log new-style health warnings to log

Signed-off-by: Sage Weil <sage@redhat.com>

mon/MDSMonitor: implement new-style cephfs health checks

Our detail elements are still strings, so we keep the bit that collapses
the metadata into a string and appends it to the string.

Each MDS-generated item becomes a detail record.

Health checks are consolidated either by MDS_ or FS_, counting the
number of mds servers or file systems affected.

Signed-off-by: Sage Weil <sage@redhat.com>

mon,mgr: pass new-style health checks from mgr's PGMap

Signed-off-by: Sage Weil <sage@redhat.com>

mon/PGMap: implement new-style health checks

Signed-off-by: Sage Weil <sage@redhat.com>

mon/OSDMonitor: implement new health checks

Signed-off-by: Sage Weil <sage@redhat.com>

mon: new health check framework

Signed-off-by: Sage Weil <sage@redhat.com>

mon: HealthMonitor -> OldHealthMonitor

This will go away post-luminous.

Signed-off-by: Sage Weil <sage@redhat.com>

mon: remove Formatter arg to QuorumService::get_health()

This is used to dump extra weirdness to the health detail structured
output, but we are about to remove all of that in luminous.

Signed-off-by: Sage Weil <sage@redhat.com>

mon/Monitor: move time sync status from 'ceph health' to 'time-sync-status' command

Signed-off-by: Sage Weil <sage@redhat.com>