]> git.apps.os.sepia.ceph.com Git - ceph.git/log
ceph.git
8 years agomon/PGMap: use int32_t, not int
Sage Weil [Fri, 19 May 2017 20:38:38 +0000 (16:38 -0400)]
mon/PGMap: use int32_t, not int

These are encoded! Be explicit.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGMap: explicitly count pg per pool
Sage Weil [Fri, 19 May 2017 20:38:14 +0000 (16:38 -0400)]
mon/PGMap: explicitly count pg per pool

This will more reliably remove empty pools from pg_pool_sum.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/MgrStatMonitor: fix version across restarts
Sage Weil [Fri, 19 May 2017 19:17:18 +0000 (15:17 -0400)]
mon/MgrStatMonitor: fix version across restarts

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGMap: cap health detail messages at 50 (configurable)
Sage Weil [Fri, 19 May 2017 15:48:15 +0000 (11:48 -0400)]
mon/PGMap: cap health detail messages at 50 (configurable)

There are two cases where we spew health detail warnings for potentially
every pg.  Cap those detail messages at 50 and, if we exceed that, include
a message saying how many more there are.  This avoids huge lists of
detail messages going from the mgr to mon and also makes life better for
users of the health detail api.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/MgrStatMonitor: trim mgrstat states
Sage Weil [Fri, 19 May 2017 15:08:26 +0000 (11:08 -0400)]
mon/MgrStatMonitor: trim mgrstat states

We don't actually need any of these older states at all so I hard-coded
a constant (oh no!).  In reality it doesn't matter what it is anyway
since PaxosService waits for paxos_service_trim_min (=250) to accumulate
before removing anything.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/MgrStatMonitor: wrap digest encoding in bufferlist
Sage Weil [Fri, 19 May 2017 15:07:34 +0000 (11:07 -0400)]
mon/MgrStatMonitor: wrap digest encoding in bufferlist

This is just for bigbang's benefit.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomgr,mon: move 'osd {scrub,deep-scrub,repair}' handling to mgr
Sage Weil [Fri, 19 May 2017 14:25:28 +0000 (10:25 -0400)]
mgr,mon: move 'osd {scrub,deep-scrub,repair}' handling to mgr

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomgr/DaemonServer: use registered osd session to send scrub messages
Sage Weil [Wed, 31 May 2017 17:02:35 +0000 (13:02 -0400)]
mgr/DaemonServer: use registered osd session to send scrub messages

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomgr/DaemonServer: keep registry of osd sessions
Sage Weil [Fri, 19 May 2017 03:06:20 +0000 (23:06 -0400)]
mgr/DaemonServer: keep registry of osd sessions

Occasionally we send them messages.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon: move parse_* helpers into cmdparse
Sage Weil [Fri, 19 May 2017 14:25:43 +0000 (10:25 -0400)]
mon: move parse_* helpers into cmdparse

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agotest/cephtool-test-mon: make ec test less sensitive to crush
Sage Weil [Fri, 19 May 2017 03:33:36 +0000 (23:33 -0400)]
test/cephtool-test-mon: make ec test less sensitive to crush

With only 3 we can get into a situation where one slot is CRUSH_ITEM_NONE.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agotest/osd/osd-dup.sh: flush_pg_stats before "ceph -s"
Kefu Chai [Fri, 19 May 2017 10:31:28 +0000 (18:31 +0800)]
test/osd/osd-dup.sh: flush_pg_stats before "ceph -s"

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agoqa/workunites/ceph-helpers.sh: move flush_pg_stats() here
Kefu Chai [Fri, 19 May 2017 10:30:49 +0000 (18:30 +0800)]
qa/workunites/ceph-helpers.sh: move flush_pg_stats() here

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agomon/PGMonitor: do not create/encode pending if luminous
Kefu Chai [Fri, 19 May 2017 09:53:27 +0000 (17:53 +0800)]
mon/PGMonitor: do not create/encode pending if luminous

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agomon/PGMonitor: reset PGMonitor member vars when luminous
Kefu Chai [Fri, 19 May 2017 09:52:52 +0000 (17:52 +0800)]
mon/PGMonitor: reset PGMonitor member vars when luminous

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agoqa/workunits/cephtool/test.sh: use jq instead of awk to select the require element
Kefu Chai [Fri, 19 May 2017 09:50:02 +0000 (17:50 +0800)]
qa/workunits/cephtool/test.sh: use jq instead of awk to select the require element

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agoqa/workunits/cephtool/test.sh: use flush_pg_stats to sync mon with osd
Kefu Chai [Fri, 19 May 2017 06:54:58 +0000 (14:54 +0800)]
qa/workunits/cephtool/test.sh: use flush_pg_stats to sync mon with osd

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agomgr: add a command "mgr report"
Kefu Chai [Wed, 17 May 2017 07:30:28 +0000 (15:30 +0800)]
mgr: add a command "mgr report"

* extract send_report() out of tick() so it can be reused.
* add a commmand "mgr report-mon" for mgr, so we are able to flush the
  the mgr stats to mon actively without waiting for the tick. this
  could help with the tests.

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agoqa/workunits/cephtool/test.sh: fix flush_pg_stats usage
Sage Weil [Thu, 18 May 2017 22:34:34 +0000 (18:34 -0400)]
qa/workunits/cephtool/test.sh: fix flush_pg_stats usage

Use a helper.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agoqa/tasks: use new reliable flush_pg_stats helper
Sage Weil [Thu, 18 May 2017 22:16:55 +0000 (18:16 -0400)]
qa/tasks: use new reliable flush_pg_stats helper

The helper gets a sequence number from the osd (or osds), and then
polls the mon until that seq is reflected there.

This is overkill in some cases, since many tests only require that the
stats be reflected on the mgr (not the mon), but waiting for it to also
reach the mon is sufficient!

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon: add 'osd last-stat-seq <osd>' command
Sage Weil [Thu, 18 May 2017 21:23:03 +0000 (17:23 -0400)]
mon: add 'osd last-stat-seq <osd>' command

Return the latest seq for the osd reflected in the mon's digest stats.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGMap: coalesce last osd stat seq in the PGMapDigest
Sage Weil [Thu, 18 May 2017 21:21:32 +0000 (17:21 -0400)]
mon/PGMap: coalesce last osd stat seq in the PGMapDigest

This is, strictly speaking, redundant, since the osd_stat is also in the
digest, but we plan to remove that.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agoosd: report a seq from flush_pg_stats command
Sage Weil [Thu, 18 May 2017 21:19:08 +0000 (17:19 -0400)]
osd: report a seq from flush_pg_stats command

Report a sequence number when we flush_pg_stats.  Combine the up_from and
a per-boot seq number to get a monotonically increasing value across OSD
restarts (we assume less than 4 billion stats reports in a single epoch).

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agoosd: include up_from, seq in osd_stat_t
Sage Weil [Thu, 18 May 2017 20:10:32 +0000 (16:10 -0400)]
osd: include up_from, seq in osd_stat_t

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGMonitor: clear PGMap data when require_luminous is set
Sage Weil [Thu, 18 May 2017 20:05:28 +0000 (16:05 -0400)]
mon/PGMonitor: clear PGMap data when require_luminous is set

Once the OSDMap flag is set there is no going back. Zero out the on-disk
PGMap data, and clear the in-memory PGMap to free up memory and make
bugs easier to spot.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agokv/RocksDBStore: make rmkeys_by_prefix efficient
Sage Weil [Thu, 18 May 2017 19:42:23 +0000 (15:42 -0400)]
kv/RocksDBStore: make rmkeys_by_prefix efficient

This matches what rm_range_keys does.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/OSDMonitor: limit number of concurrently creating pgs
Sage Weil [Thu, 18 May 2017 19:41:39 +0000 (15:41 -0400)]
mon/OSDMonitor: limit number of concurrently creating pgs

There is overhead for PGs we are creating because the mon has to track
which OSD each one current maps to.  This can be problematic on a very
large cluster.  Limit the overhead by setting a cap on the number of PGs
we are creating at once; leave the rest in a persistent queue.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/MgrStatMonitor: fix digest vs pending digest
Sage Weil [Thu, 18 May 2017 19:32:57 +0000 (15:32 -0400)]
mon/MgrStatMonitor: fix digest vs pending digest

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/OSDMonitor: fix bad update_pending_pgs call
Sage Weil [Thu, 18 May 2017 19:32:26 +0000 (15:32 -0400)]
mon/OSDMonitor: fix bad update_pending_pgs call

We are not persisiting the updated creating_pgs here; this is bad!  I'm
not sure why it was there to begin with?

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGMonitor: disable when REQUIRE_LUMINOUS is set
Sage Weil [Thu, 18 May 2017 17:57:58 +0000 (13:57 -0400)]
mon/PGMonitor: disable when REQUIRE_LUMINOUS is set

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agotest: pass mon_pg_warn_min_per_osd=3 to mgr also
Kefu Chai [Thu, 18 May 2017 14:24:23 +0000 (22:24 +0800)]
test: pass mon_pg_warn_min_per_osd=3 to mgr also

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agomon: move dump_info() to PGStatService
Kefu Chai [Thu, 18 May 2017 11:58:50 +0000 (19:58 +0800)]
mon: move dump_info() to PGStatService

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agomon: more constness to PGStatService
Kefu Chai [Thu, 18 May 2017 11:50:36 +0000 (19:50 +0800)]
mon: more constness to PGStatService

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agomon/OSDMonitor: do not reference pgservice if quorum not formed
Kefu Chai [Wed, 17 May 2017 10:18:05 +0000 (18:18 +0800)]
mon/OSDMonitor: do not reference pgservice if quorum not formed

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agotest/cephtool-test-*: enable mgr
Sage Weil [Wed, 17 May 2017 14:14:28 +0000 (10:14 -0400)]
test/cephtool-test-*: enable mgr

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGMap: add to mempool
Sage Weil [Tue, 16 May 2017 22:21:19 +0000 (18:21 -0400)]
mon/PGMap: add to mempool

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGMap: use auto
Sage Weil [Tue, 16 May 2017 21:52:03 +0000 (17:52 -0400)]
mon/PGMap: use auto

I'm not rewriting this to use range iterator syntax because I'm in a
hurry. This just lets me change the types without touching all this code
again.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agoqa/workunites/ceph-helpers.sh: do not bail when num_pg==0
Sage Weil [Tue, 16 May 2017 21:25:04 +0000 (17:25 -0400)]
qa/workunites/ceph-helpers.sh: do not bail when num_pg==0

Right after the cluster is created when the first mgr report hasn't come
in yet we will report 0 pgs.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agotest/osd/osd-dup.sh: run mgr with test
Sage Weil [Tue, 16 May 2017 21:24:23 +0000 (17:24 -0400)]
test/osd/osd-dup.sh: run mgr with test

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomgr/ClusterState: report health about pgs
Sage Weil [Tue, 16 May 2017 17:27:00 +0000 (13:27 -0400)]
mgr/ClusterState: report health about pgs

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGMonitor: move most of get_health into PGMap
Sage Weil [Tue, 16 May 2017 17:20:51 +0000 (13:20 -0400)]
mon/PGMonitor: move most of get_health into PGMap

Leave the legacy full/nearfull stuff in PGMonitor (although we'll still
need to clean out the PGMap fields too).

This is cut and paste, some stripping of pg_map., g_conf -> cct->_conf,
and a bit of whitespace cleanup.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon: accept health summary and detail messages from mgr
Sage Weil [Tue, 16 May 2017 17:17:17 +0000 (13:17 -0400)]
mon: accept health summary and detail messages from mgr

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGMapStatService: various updates
Kefu Chai [Tue, 16 May 2017 07:30:16 +0000 (15:30 +0800)]
mon/PGMapStatService: various updates

* add override specifier
* s/parent/pgmap/
* remove unnecessary forward declarations
* make `pg_map` and `pgservice` private member variables.
* formatting changes: wrap at 80, for example.
* define maybe_add_creating_pgs() instead of get_pg_stat()
* use unique_ptr<PGMapStatService> for holding the pointer.

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agomon/PGStatService: define maybe_trim_creating_pgs() method
Kefu Chai [Tue, 16 May 2017 07:26:41 +0000 (15:26 +0800)]
mon/PGStatService: define maybe_trim_creating_pgs() method

replace get_pg_stat() with maybe_trim_creating_pgs(), so it is more
symmetric: maybe_add_creating_pgs() versus maybe_trim_creating_pgs().
and we don't need to return a pointer which is always valid. a const
reference is a better fit in this case.

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agomon/MgrStatMonitor: add override specifier for virtual methods
Kefu Chai [Tue, 16 May 2017 07:22:34 +0000 (15:22 +0800)]
mon/MgrStatMonitor: add override specifier for virtual methods

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agoosd: stop sending pgstat if luminous
Kefu Chai [Tue, 16 May 2017 05:52:51 +0000 (13:52 +0800)]
osd: stop sending pgstat if luminous

we don't send pgstat to mon from osd once we set
CEPH_OSDMAP_REQUIRE_LUMINOUS. and cleanup the pgstat queue when we
do the switch.

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agomon/PGMonitor: reindent maybe_add_creating_pgs()
Kefu Chai [Tue, 16 May 2017 05:51:48 +0000 (13:51 +0800)]
mon/PGMonitor: reindent maybe_add_creating_pgs()

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agomon: move mgr report handling into separate MgrStatMonitor
Sage Weil [Tue, 16 May 2017 16:44:54 +0000 (12:44 -0400)]
mon: move mgr report handling into separate MgrStatMonitor

We don't want this is MgrMonitor or else the MgrMap epoch will roll over
every few seconds when we get a mgr report.  Also, it's nice to keep this
service separate from the unrelated MgrMap management.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGMonitor: link PGStatService to the actual PGMap
Sage Weil [Mon, 15 May 2017 21:45:06 +0000 (17:45 -0400)]
mon/PGMonitor: link PGStatService to the actual PGMap

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon: adjust pgservice based on OSDMap require_luminous flag
Sage Weil [Mon, 15 May 2017 18:53:57 +0000 (14:53 -0400)]
mon: adjust pgservice based on OSDMap require_luminous flag

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon: remove last trace of PGMap from MgrMonitor
Kefu Chai [Mon, 15 May 2017 13:46:32 +0000 (21:46 +0800)]
mon: remove last trace of PGMap from MgrMonitor

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agomgr: send PGMapDigest instead of PGMap to mon
Sage Weil [Sat, 13 May 2017 04:51:45 +0000 (23:51 -0500)]
mgr: send PGMapDigest instead of PGMap to mon

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/OSDMonitor: avoid pulling pgmap ratio on mkfs
Sage Weil [Sat, 13 May 2017 02:40:22 +0000 (21:40 -0500)]
mon/OSDMonitor: avoid pulling pgmap ratio on mkfs

If this is the first osdmap epoch there is nothing to be learned from
the PGMap.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGStatService: implement pre-luminous methods in parent
Sage Weil [Sat, 13 May 2017 02:30:43 +0000 (21:30 -0500)]
mon/PGStatService: implement pre-luminous methods in parent

Many of these are only needed to support PGMonitor.  Implement those
in the parent and make them ceph_abort() so that the new child can avoid
the clutter.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGMap: move summary information into parent PGMapDigest object
Sage Weil [Sat, 13 May 2017 02:27:11 +0000 (21:27 -0500)]
mon/PGMap: move summary information into parent PGMapDigest object

Everything summary-ish that we need to send to the mon is moved into a
parent class.  The child PGMap retains the detail.

The parent gets its own encode(), and PGMap::encode_digest() will call it
to encode just the summary info.

Squashed in here is a new num_pg_by_osd that could have been done
in a preceding patch but I did things in the wrong order.  :(

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/OSDMonitor: condition PGMap->creating priming on !luminous
Sage Weil [Sat, 13 May 2017 02:08:57 +0000 (21:08 -0500)]
mon/OSDMonitor: condition PGMap->creating priming on !luminous

Once we set the luminous flag we shouldn't need this anymore.  In fact,
we should stop needing this immediately after we set the monitor
luminous flag.

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGMap: use num_pg where available
Sage Weil [Sat, 13 May 2017 01:45:29 +0000 (20:45 -0500)]
mon/PGMap: use num_pg where available

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon/PGMap: remove version from summary
Sage Weil [Sat, 13 May 2017 01:41:01 +0000 (20:41 -0500)]
mon/PGMap: remove version from summary

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agomon: give PGMonitor and MgrMonitor their own PGStatService implementations
Greg Farnum [Tue, 9 May 2017 20:06:56 +0000 (13:06 -0700)]
mon: give PGMonitor and MgrMonitor their own PGStatService implementations

Right now they pretty much just copy the PGMapStatService, but it'll
let us slim down the MgrMonitor's one more easily.

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: move the PGStatService lifecycle into the MgrMon/PGMon
Greg Farnum [Wed, 15 Mar 2017 22:24:51 +0000 (15:24 -0700)]
mon: move the PGStatService lifecycle into the MgrMon/PGMon

I remain lazy enough that we just use the MgrMon right now for easier
testing.

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: pgstats: move some creating_pgs maintenance into the PGStatService
Greg Farnum [Tue, 9 May 2017 19:01:22 +0000 (12:01 -0700)]
mon: pgstats: move some creating_pgs maintenance into the PGStatService

And remove get_pg_map() now that there are no users!

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: pgstat: remove unneeded StatIter functions
Greg Farnum [Tue, 9 May 2017 18:12:29 +0000 (11:12 -0700)]
mon: pgstat: remove unneeded StatIter functions

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: pgstat: expose the full pg_stat map to remove a get_pg_map() call
Greg Farnum [Fri, 5 May 2017 20:37:21 +0000 (13:37 -0700)]
mon: pgstat: expose the full pg_stat map to remove a get_pg_map() call

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: pgstat: expose the full osd_stat map to remove a get_pg_map() call
Greg Farnum [Fri, 5 May 2017 20:16:48 +0000 (13:16 -0700)]
mon: pgstat: expose the full osd_stat map to remove a get_pg_map() call

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: subclass PGStatService into PGMapStatService
Greg Farnum [Fri, 5 May 2017 20:48:02 +0000 (13:48 -0700)]
mon: subclass PGStatService into PGMapStatService

Here we begin moving chunks into more appropriate locations.

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: give PGStatService a reweight_by_utilization indirection
Greg Farnum [Fri, 5 May 2017 04:49:17 +0000 (21:49 -0700)]
mon: give PGStatService a reweight_by_utilization indirection

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: give PGStatService a process_pg_command indirection
Greg Farnum [Fri, 5 May 2017 04:42:24 +0000 (21:42 -0700)]
mon: give PGStatService a process_pg_command indirection

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: use a pointer to PGStatService instead of putting it in Monitor
Greg Farnum [Wed, 15 Mar 2017 21:33:33 +0000 (14:33 -0700)]
mon: use a pointer to PGStatService instead of putting it in Monitor

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: kill direct references to pgmon from OSDMonitor
Greg Farnum [Thu, 4 May 2017 19:45:25 +0000 (12:45 -0700)]
mon: kill direct references to pgmon from OSDMonitor

The only one left now is where the OSDMonitor triggers
PGMonitor::check_osd_map() and that's going to die post-Luminous,
so no need to do anything fancy with it.

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: get the OSDMonitor's pg info dump functions away from using PGMap
Greg Farnum [Wed, 15 Mar 2017 20:47:39 +0000 (13:47 -0700)]
mon: get the OSDMonitor's pg info dump functions away from using PGMap

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: stop dereferencing pg_map for object stats in OSDMonitor
Greg Farnum [Wed, 15 Mar 2017 00:46:48 +0000 (17:46 -0700)]
mon: stop dereferencing pg_map for object stats in OSDMonitor

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: yet more OSDMonitor dep-pgmap-ification
Greg Farnum [Wed, 15 Mar 2017 00:18:36 +0000 (17:18 -0700)]
mon: yet more OSDMonitor dep-pgmap-ification

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: more OSDMonitor de-pgmap-ification
Greg Farnum [Tue, 14 Mar 2017 23:29:27 +0000 (16:29 -0700)]
mon: more OSDMonitor de-pgmap-ification

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: OSDMonitor starting to abstract PGMap via PGService
Greg Farnum [Fri, 17 Mar 2017 22:42:38 +0000 (15:42 -0700)]
mon: OSDMonitor starting to abstract PGMap via PGService

We added an unfortunate get_pg_map() function but it's still easy
to find and puts off some of the logic work. We'll get to it soon.

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: FSCommands gets pool stats via PGService
Greg Farnum [Fri, 17 Mar 2017 22:41:14 +0000 (15:41 -0700)]
mon: FSCommands gets pool stats via PGService

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: FSCommands gets pool stats via PGService
Greg Farnum [Tue, 14 Mar 2017 23:11:04 +0000 (16:11 -0700)]
mon: FSCommands gets pool stats via PGService

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: outline a PGStatService to replace raw PGMap calls
Greg Farnum [Tue, 14 Mar 2017 22:43:08 +0000 (15:43 -0700)]
mon: outline a PGStatService to replace raw PGMap calls

Right now, this is a quick hack that derives from PGMap. Soon, it will
be more interesting.

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: mgr: send a pgmap in reports
Greg Farnum [Sat, 11 Mar 2017 03:09:22 +0000 (19:09 -0800)]
mon: mgr: send a pgmap in reports

This definitely isn't appropriate for long term but is a quick route to testing

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon: route manager reports down to the MgrMon
Greg Farnum [Sat, 11 Mar 2017 02:25:13 +0000 (18:25 -0800)]
mon: route manager reports down to the MgrMon

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomgr: introduce tick() calls through the system
Greg Farnum [Sat, 11 Mar 2017 02:48:47 +0000 (18:48 -0800)]
mgr: introduce tick() calls through the system

Still empty for the moment

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomgr: create a skeleton MMonMgrReport message type
Greg Farnum [Sat, 11 Mar 2017 02:21:17 +0000 (18:21 -0800)]
mgr: create a skeleton MMonMgrReport message type

We'll use this for sending data to the monitor at tick intervals.

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomgr: switch from send_beacon() to tick()
Greg Farnum [Fri, 10 Mar 2017 23:02:50 +0000 (15:02 -0800)]
mgr: switch from send_beacon() to tick()

We are about to want similar time-based stuff in our other
classes and it's simplest to coalesce them into a single wakeup.

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agoMonClient: support "passthrough" of MonMap messages so other Dispatchers can see...
Greg Farnum [Tue, 7 Mar 2017 23:30:55 +0000 (15:30 -0800)]
MonClient: support "passthrough" of MonMap messages so other Dispatchers can see them

Use this in the Manager so it can notify of new monitor maps.

Signed-off-by: Greg Farnum <gfarnum@redhat.com>
8 years agomon/MgrMap: prefix summary with epoch; add comma
Sage Weil [Tue, 16 May 2017 22:40:08 +0000 (18:40 -0400)]
mon/MgrMap: prefix summary with epoch; add comma

This matches the others now:

    cluster 0d553373-5de6-4d90-bc60-cfd6f6554631
     health HEALTH_OK
     monmap e2: 1 mons, quorum a
      fsmap e2: 0/0/1 up
        mgr e3: active: x(starting), standbys: y
     osdmap e5: 1 osds: 0 up, 0 in
      pgmap v6: 24 pgs, 3 pools, 0 bytes data, 0 objects
            0 kB used, 0 kB / 0 kB avail
            100.000% pgs inactive
                  24 creating

Signed-off-by: Sage Weil <sage@redhat.com>
8 years agoMerge pull request #15427 from tchaikov/wip-vstart-osd-id
Sage Weil [Fri, 2 Jun 2017 15:34:45 +0000 (10:34 -0500)]
Merge pull request #15427 from tchaikov/wip-vstart-osd-id

vstart: print "start osd.$id" instead of "start osd$id"

Reviewed-by: Sage Weil <sage@redhat.com>
8 years agoMerge pull request #15217 from badone/wip-async-sleep-timer-fix
Sage Weil [Fri, 2 Jun 2017 15:32:15 +0000 (10:32 -0500)]
Merge pull request #15217 from badone/wip-async-sleep-timer-fix

osd: Move scrub sleep timer to osdservice

Reviewed-by: Josh Durgin <jdurgin@redhat.com>
8 years agoMerge pull request #15396 from jcsp/wip-ceph-status
Sage Weil [Fri, 2 Jun 2017 15:30:55 +0000 (10:30 -0500)]
Merge pull request #15396 from jcsp/wip-ceph-status

mon: revise "ceph status" output

Reviewed-by: Sage Weil <sage@redhat.com>
8 years agoMerge pull request #15011 from xiexingguo/wip-crush-tool
Sage Weil [Fri, 2 Jun 2017 15:20:44 +0000 (10:20 -0500)]
Merge pull request #15011 from xiexingguo/wip-crush-tool

tools/crushtool: replicated-rule API support

Reviewed-by: Sage Weil <sage@redhat.com>
8 years agoMerge pull request #15176 from SUSE/wip-mgr-fix-help
Sage Weil [Fri, 2 Jun 2017 15:20:27 +0000 (10:20 -0500)]
Merge pull request #15176 from SUSE/wip-mgr-fix-help

mgr,osd:  ceph-mgr --help, unify usage text of other daemons

Reviewed-by: Kefu Chai <kchai@redhat.com>
8 years agoMerge pull request #15395 from liewegas/wip-log-entity
Sage Weil [Fri, 2 Jun 2017 15:19:58 +0000 (10:19 -0500)]
Merge pull request #15395 from liewegas/wip-log-entity

common/LogEntry: include EntityName in log entries

Reviewed-by: Kefu Chai <kchai@redhat.com>
8 years agoMerge pull request #15398 from liewegas/wip-bluestore-leak
Sage Weil [Fri, 2 Jun 2017 15:19:26 +0000 (10:19 -0500)]
Merge pull request #15398 from liewegas/wip-bluestore-leak

os/bluestore: fix buffers pinned by indefinitely deferred writes

Reviewed-by: Igor Fedotov <ifedotov@mirantis.com>
8 years agoMerge pull request #15402 from liewegas/wip-bluestore-mempools
Sage Weil [Fri, 2 Jun 2017 15:19:03 +0000 (10:19 -0500)]
Merge pull request #15402 from liewegas/wip-bluestore-mempools

os/bluestore: improve mempool usage

Reviewed-by: Igor Fedotov <ifedotov@mirantis.com>
8 years agoMerge pull request #15403 from liewegas/wip-mempool-dump
Sage Weil [Fri, 2 Jun 2017 15:18:38 +0000 (10:18 -0500)]
Merge pull request #15403 from liewegas/wip-mempool-dump

mempool: improve dump; fix buffer accounting bugs

Reviewed-by: Kefu Chai <kchai@redhat.com>
8 years agoMerge pull request #15431 from rzarzynski/wip-rgw-vim-modeline-for-orhpantool
Casey Bodley [Fri, 2 Jun 2017 14:05:17 +0000 (10:05 -0400)]
Merge pull request #15431 from rzarzynski/wip-rgw-vim-modeline-for-orhpantool

rgw: add the Vim's modeline into rgw_orphan.cc.

Reviewed-by: Casey Bodley <cbodley@redhat.com>
8 years agoMerge pull request #15425 from yaozongyou/fix-build-error
Kefu Chai [Fri, 2 Jun 2017 13:56:03 +0000 (21:56 +0800)]
Merge pull request #15425 from yaozongyou/fix-build-error

cmake: link against fcgi only if enabled

Reviewed-by: Casey Bodley <cbodley@redhat.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>
8 years agorgw: add the Vim's modeline into rgw_orphan.cc. 15431/head
Radoslaw Zarzynski [Fri, 2 Jun 2017 12:03:44 +0000 (14:03 +0200)]
rgw: add the Vim's modeline into rgw_orphan.cc.

Signed-off-by: Radoslaw Zarzynski <rzarzynski@mirantis.com>
8 years agovstart: print "start osd.$id" instead of "start osd$id" 15427/head
Kefu Chai [Fri, 2 Jun 2017 07:10:38 +0000 (15:10 +0800)]
vstart: print "start osd.$id" instead of "start osd$id"

"osd.9" is more consistent with other places where an osd is referenced.

Signed-off-by: Kefu Chai <kchai@redhat.com>
8 years agoMerge pull request #15408 from yonghengdexin735/wip-zzz-add_const
Jos Collin [Fri, 2 Jun 2017 05:41:58 +0000 (11:11 +0530)]
Merge pull request #15408 from yonghengdexin735/wip-zzz-add_const

include/lru.h: add const to member functions

Reviewed-by: Jos Collin <jcollin@redhat.com>
8 years agoFix cmake build error. 15425/head
Yao Zongyou [Fri, 2 Jun 2017 04:11:59 +0000 (12:11 +0800)]
Fix cmake build error.

Signed-off-by: Yao Zongyou <yaozongyou@vip.qq.com>
8 years agoMerge pull request #15420 from yuriw/wip_remove_default_idle_timeout
Sage Weil [Fri, 2 Jun 2017 03:08:31 +0000 (22:08 -0500)]
Merge pull request #15420 from yuriw/wip_remove_default_idle_timeout

qa: Removed all 'default_idle_timeout' due to chnage in rwg task