1 // -*- mode:C++; tab-width:8; c-basic-offset:2; indent-tabs-mode:t -*-
2 // vim: ts=8 sw=2 smarttab
4 * Ceph - scalable distributed file system
6 * Copyright (C) 2004-2006 Sage Weil <sage@newdream.net>
7 * Copyright (C) 2013,2014 Cloudwatt <libre.licensing@cloudwatt.com>
9 * Author: Loic Dachary <loic@dachary.org>
11 * This is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU Lesser General Public
13 * License version 2.1, as published by the Free Software
14 * Foundation. See file COPYING.
23 * describe properties of the OSD cluster.
24 * disks, disk groups, total # osds,
33 #include <boost/smart_ptr/local_shared_ptr.hpp>
34 #include "include/btree_map.h"
35 #include "include/common_fwd.h"
36 #include "include/types.h"
37 #include "common/ceph_releases.h"
38 #include "osd_types.h"
40 //#include "include/ceph_features.h"
41 #include "crush/CrushWrapper.h"
43 // forward declaration
45 class health_check_map_t
;
48 * we track up to two intervals during which the osd was alive and
49 * healthy. the most recent is [up_from,up_thru), where up_thru is
50 * the last epoch the osd is known to have _started_. i.e., a lower
51 * bound on the actual osd death. down_at (if it is > up_from) is an
52 * upper bound on the actual osd death.
54 * the second is the last_clean interval [begin,end). in that case,
55 * the last interval is the last epoch known to have been either
56 * _finished_, or during which the osd cleanly shut down. when
57 * possible, we push this forward to the epoch the osd was eventually
60 * the lost_at is used to allow build_prior to proceed without waiting
61 * for an osd to recover. In certain cases, progress may be blocked
62 * because an osd is down that may contain updates (i.e., a pg may have
63 * gone rw during an interval). If the osd can't be brought online, we
64 * can force things to proceed knowing that we _might_ be losing some
65 * acked writes. If the osd comes back to life later, that's fine to,
66 * but those writes will still be lost (the divergent objects will be
70 epoch_t last_clean_begin
; // last interval that ended with a clean osd shutdown
71 epoch_t last_clean_end
;
72 epoch_t up_from
; // epoch osd marked up
73 epoch_t up_thru
; // lower bound on actual osd death (if > up_from)
74 epoch_t down_at
; // upper bound on actual osd death (if > up_from)
75 epoch_t lost_at
; // last epoch we decided data was "lost"
77 osd_info_t() : last_clean_begin(0), last_clean_end(0),
78 up_from(0), up_thru(0), down_at(0), lost_at(0) {}
80 void dump(ceph::Formatter
*f
) const;
81 void encode(ceph::buffer::list
& bl
) const;
82 void decode(ceph::buffer::list::const_iterator
& bl
);
83 static void generate_test_instances(std::list
<osd_info_t
*>& o
);
85 WRITE_CLASS_ENCODER(osd_info_t
)
87 std::ostream
& operator<<(std::ostream
& out
, const osd_info_t
& info
);
90 utime_t down_stamp
; ///< timestamp when we were last marked down
91 float laggy_probability
; ///< encoded as __u32: 0 = definitely not laggy, 0xffffffff definitely laggy
92 __u32 laggy_interval
; ///< average interval between being marked laggy and recovering
93 uint64_t features
; ///< features supported by this osd we should know about
94 __u32 old_weight
; ///< weight prior to being auto marked out
95 utime_t last_purged_snaps_scrub
; ///< last scrub of purged_snaps
96 epoch_t dead_epoch
= 0; ///< last epoch we were confirmed dead (not just down)
98 osd_xinfo_t() : laggy_probability(0), laggy_interval(0),
99 features(0), old_weight(0) {}
101 void dump(ceph::Formatter
*f
) const;
102 void encode(ceph::buffer::list
& bl
, uint64_t features
) const;
103 void decode(ceph::buffer::list::const_iterator
& bl
);
104 static void generate_test_instances(std::list
<osd_xinfo_t
*>& o
);
106 WRITE_CLASS_ENCODER_FEATURES(osd_xinfo_t
)
108 std::ostream
& operator<<(std::ostream
& out
, const osd_xinfo_t
& xi
);
113 ceph::buffer::list data
;
114 typedef btree::btree_map
<pg_t
,ceph_le32
*> map_t
;
117 void encode(ceph::buffer::list
& bl
) const {
119 uint32_t n
= map
.size();
121 for (auto &p
: map
) {
123 bl
.append((char*)p
.second
, (*p
.second
+ 1) * sizeof(ceph_le32
));
126 void decode(ceph::buffer::list::const_iterator
& p
) {
135 size_t start_off
= pstart
.get_off();
136 std::vector
<std::pair
<pg_t
,size_t>> offsets
;
138 for (unsigned i
=0; i
<n
; ++i
) {
141 offsets
[i
].first
= pgid
;
142 offsets
[i
].second
= p
.get_off() - start_off
;
145 p
+= vn
* sizeof(int32_t);
147 size_t len
= p
.get_off() - start_off
;
148 pstart
.copy(len
, data
);
149 if (data
.get_num_buffers() > 1) {
153 char *start
= data
.c_str();
154 for (auto i
: offsets
) {
155 map
.insert(map
.end(), std::make_pair(i
.first
, (ceph_le32
*)(start
+ i
.second
)));
159 ceph::buffer::list bl
;
161 auto p
= std::cbegin(bl
);
164 friend bool operator==(const PGTempMap
& l
, const PGTempMap
& r
) {
166 l
.map
.size() == r
.map
.size() &&
167 l
.data
.contents_equal(r
.data
);
171 map_t::const_iterator it
;
172 map_t::const_iterator end
;
173 std::pair
<pg_t
,std::vector
<int32_t>> current
;
174 void init_current() {
176 current
.first
= it
->first
;
177 ceph_assert(it
->second
);
178 current
.second
.resize(*it
->second
);
179 ceph_le32
*p
= it
->second
+ 1;
180 for (uint32_t n
= 0; n
< *it
->second
; ++n
, ++p
) {
181 current
.second
[n
] = *p
;
186 iterator(map_t::const_iterator p
,
187 map_t::const_iterator e
)
192 const std::pair
<pg_t
,std::vector
<int32_t>>& operator*() const {
195 const std::pair
<pg_t
,std::vector
<int32_t>>* operator->() const {
198 friend bool operator==(const iterator
& l
, const iterator
& r
) {
201 friend bool operator!=(const iterator
& l
, const iterator
& r
) {
204 iterator
& operator++() {
210 iterator
operator++(int) {
218 iterator
begin() const {
219 return iterator(map
.begin(), map
.end());
221 iterator
end() const {
222 return iterator(map
.end(), map
.end());
224 iterator
find(pg_t pgid
) const {
225 return iterator(map
.find(pgid
), map
.end());
227 size_t size() const {
230 size_t count(pg_t pgid
) const {
231 return map
.count(pgid
);
233 void erase(pg_t pgid
) {
240 void set(pg_t pgid
, const mempool::osdmap::vector
<int32_t>& v
) {
242 size_t need
= sizeof(ceph_le32
) * (1 + v
.size());
243 if (need
< data
.get_append_buffer_unused_tail_length()) {
244 ceph::buffer::ptr
z(data
.get_append_buffer_unused_tail_length());
246 data
.append(z
.c_str(), z
.length());
249 map
[pgid
] = (ceph_le32
*)(data
.back().end_c_str()) - (1 + v
.size());
251 mempool::osdmap::vector
<int32_t> get(pg_t pgid
) {
252 mempool::osdmap::vector
<int32_t> v
;
253 ceph_le32
*p
= map
[pgid
];
256 for (size_t i
= 0; i
< n
; ++i
, ++p
) {
262 // trivial implementation
263 mempool::osdmap::map
<pg_t
,mempool::osdmap::vector
<int32_t> > pg_temp
;
265 void encode(ceph::buffer::list
& bl
) const {
268 void decode(ceph::buffer::list::const_iterator
& p
) {
271 friend bool operator==(const PGTempMap
& l
, const PGTempMap
& r
) {
273 l
.pg_temp
.size() == r
.pg_temp
.size() &&
274 l
.pg_temp
== r
.pg_temp
;
278 mempool::osdmap::map
<pg_t
,mempool::osdmap::vector
<int32_t> >::const_iterator it
;
280 iterator(mempool::osdmap::map
<pg_t
,
281 mempool::osdmap::vector
<int32_t> >::const_iterator p
)
284 std::pair
<pg_t
,const mempool::osdmap::vector
<int32_t>&> operator*() const {
287 const std::pair
<const pg_t
,mempool::osdmap::vector
<int32_t>>* operator->() const {
290 friend bool operator==(const iterator
& l
, const iterator
& r
) {
293 friend bool operator!=(const iterator
& l
, const iterator
& r
) {
296 iterator
& operator++() {
300 iterator
operator++(int) {
306 iterator
begin() const {
307 return iterator(pg_temp
.cbegin());
309 iterator
end() const {
310 return iterator(pg_temp
.cend());
312 iterator
find(pg_t pgid
) const {
313 return iterator(pg_temp
.find(pgid
));
315 size_t size() const {
316 return pg_temp
.size();
318 size_t count(pg_t pgid
) const {
319 return pg_temp
.count(pgid
);
321 void erase(pg_t pgid
) {
327 void set(pg_t pgid
, const mempool::osdmap::vector
<int32_t>& v
) {
330 const mempool::osdmap::vector
<int32_t>& get(pg_t pgid
) {
331 return pg_temp
.at(pgid
);
334 void dump(ceph::Formatter
*f
) const {
335 for (const auto &pg
: *this) {
336 f
->open_object_section("osds");
337 f
->dump_stream("pgid") << pg
.first
;
338 f
->open_array_section("osds");
339 for (const auto osd
: pg
.second
)
340 f
->dump_int("osd", osd
);
346 WRITE_CLASS_ENCODER(PGTempMap
)
352 MEMPOOL_CLASS_HELPERS();
356 MEMPOOL_CLASS_HELPERS();
358 /// feature bits we were encoded with. the subsequent OSDMap
359 /// encoding should match.
360 uint64_t encode_features
;
362 epoch_t epoch
; // new epoch; we are a diff from epoch-1 to epoch
364 int64_t new_pool_max
; //incremented by the OSDMonitor on each pool create
366 ceph_release_t new_require_osd_release
{0xff};
367 uint32_t new_stretch_bucket_count
{0};
368 uint32_t new_degraded_stretch_mode
{0};
369 uint32_t new_recovering_stretch_mode
{0};
370 int32_t new_stretch_mode_bucket
{0};
371 bool stretch_mode_enabled
{false};
372 bool change_stretch_mode
{false};
375 ceph::buffer::list fullmap
; // in lieu of below.
376 ceph::buffer::list crush
;
380 mempool::osdmap::map
<int64_t,pg_pool_t
> new_pools
;
381 mempool::osdmap::map
<int64_t,std::string
> new_pool_names
;
382 mempool::osdmap::set
<int64_t> old_pools
;
383 mempool::osdmap::map
<std::string
,std::map
<std::string
,std::string
> > new_erasure_code_profiles
;
384 mempool::osdmap::vector
<std::string
> old_erasure_code_profiles
;
385 mempool::osdmap::map
<int32_t,entity_addrvec_t
> new_up_client
;
386 mempool::osdmap::map
<int32_t,entity_addrvec_t
> new_up_cluster
;
387 mempool::osdmap::map
<int32_t,uint32_t> new_state
; // XORed onto previous state.
388 mempool::osdmap::map
<int32_t,uint32_t> new_weight
;
389 mempool::osdmap::map
<pg_t
,mempool::osdmap::vector
<int32_t> > new_pg_temp
; // [] to remove
390 mempool::osdmap::map
<pg_t
, int32_t> new_primary_temp
; // [-1] to remove
391 mempool::osdmap::map
<int32_t,uint32_t> new_primary_affinity
;
392 mempool::osdmap::map
<int32_t,epoch_t
> new_up_thru
;
393 mempool::osdmap::map
<int32_t,std::pair
<epoch_t
,epoch_t
> > new_last_clean_interval
;
394 mempool::osdmap::map
<int32_t,epoch_t
> new_lost
;
395 mempool::osdmap::map
<int32_t,uuid_d
> new_uuid
;
396 mempool::osdmap::map
<int32_t,osd_xinfo_t
> new_xinfo
;
398 mempool::osdmap::map
<entity_addr_t
,utime_t
> new_blocklist
;
399 mempool::osdmap::vector
<entity_addr_t
> old_blocklist
;
400 mempool::osdmap::map
<int32_t, entity_addrvec_t
> new_hb_back_up
;
401 mempool::osdmap::map
<int32_t, entity_addrvec_t
> new_hb_front_up
;
403 mempool::osdmap::map
<pg_t
,mempool::osdmap::vector
<int32_t>> new_pg_upmap
;
404 mempool::osdmap::map
<pg_t
,mempool::osdmap::vector
<std::pair
<int32_t,int32_t>>> new_pg_upmap_items
;
405 mempool::osdmap::set
<pg_t
> old_pg_upmap
, old_pg_upmap_items
;
406 mempool::osdmap::map
<int64_t, snap_interval_set_t
> new_removed_snaps
;
407 mempool::osdmap::map
<int64_t, snap_interval_set_t
> new_purged_snaps
;
409 mempool::osdmap::map
<int32_t,uint32_t> new_crush_node_flags
;
410 mempool::osdmap::map
<int32_t,uint32_t> new_device_class_flags
;
412 std::string cluster_snapshot
;
414 float new_nearfull_ratio
= -1;
415 float new_backfillfull_ratio
= -1;
416 float new_full_ratio
= -1;
418 ceph_release_t new_require_min_compat_client
{0xff};
420 utime_t new_last_up_change
, new_last_in_change
;
422 mutable bool have_crc
; ///< crc values are defined
423 uint32_t full_crc
; ///< crc of the resulting OSDMap
424 mutable uint32_t inc_crc
; ///< crc of this incremental
426 int get_net_marked_out(const OSDMap
*previous
) const;
427 int get_net_marked_down(const OSDMap
*previous
) const;
428 int identify_osd(uuid_d u
) const;
430 void encode_client_old(ceph::buffer::list
& bl
) const;
431 void encode_classic(ceph::buffer::list
& bl
, uint64_t features
) const;
432 void encode(ceph::buffer::list
& bl
, uint64_t features
=CEPH_FEATURES_ALL
) const;
433 void decode_classic(ceph::buffer::list::const_iterator
&p
);
434 void decode(ceph::buffer::list::const_iterator
&bl
);
435 void dump(ceph::Formatter
*f
) const;
436 static void generate_test_instances(std::list
<Incremental
*>& o
);
438 explicit Incremental(epoch_t e
=0) :
440 epoch(e
), new_pool_max(-1), new_flags(-1), new_max_osd(-1),
441 have_crc(false), full_crc(0), inc_crc(0) {
443 explicit Incremental(ceph::buffer::list
&bl
) {
444 auto p
= std::cbegin(bl
);
447 explicit Incremental(ceph::buffer::list::const_iterator
&p
) {
451 pg_pool_t
*get_new_pool(int64_t pool
, const pg_pool_t
*orig
) {
452 if (new_pools
.count(pool
) == 0)
453 new_pools
[pool
] = *orig
;
454 return &new_pools
[pool
];
456 bool has_erasure_code_profile(const std::string
&name
) const {
457 auto i
= new_erasure_code_profiles
.find(name
);
458 return i
!= new_erasure_code_profiles
.end();
460 void set_erasure_code_profile(const std::string
&name
,
461 const std::map
<std::string
,std::string
>& profile
) {
462 new_erasure_code_profiles
[name
] = profile
;
464 mempool::osdmap::map
<std::string
,std::map
<std::string
,std::string
>> get_erasure_code_profiles() const {
465 return new_erasure_code_profiles
;
468 /// propagate update pools' (snap and other) metadata to any of their tiers
469 int propagate_base_properties_to_tiers(CephContext
*cct
, const OSDMap
&base
);
471 /// filter out osds with any pending state changing
472 size_t get_pending_state_osds(std::vector
<int> *osds
) {
476 for (auto &p
: new_state
) {
477 osds
->push_back(p
.first
);
483 bool pending_osd_has_state(int osd
, unsigned state
) {
484 return new_state
.count(osd
) && (new_state
[osd
] & state
) != 0;
487 bool pending_osd_state_set(int osd
, unsigned state
) {
488 if (pending_osd_has_state(osd
, state
))
490 new_state
[osd
] |= state
;
494 // cancel the specified pending osd state if there is any
495 // return ture on success, false otherwise.
496 bool pending_osd_state_clear(int osd
, unsigned state
) {
497 if (!pending_osd_has_state(osd
, state
)) {
498 // never has been set or already has been cancelled.
502 new_state
[osd
] &= ~state
;
503 if (!new_state
[osd
]) {
505 new_state
.erase(osd
);
510 bool in_new_removed_snaps(int64_t pool
, snapid_t snap
) const {
511 auto p
= new_removed_snaps
.find(pool
);
512 if (p
== new_removed_snaps
.end()) {
515 return p
->second
.contains(snap
);
521 epoch_t epoch
; // what epoch of the osd cluster descriptor is this
522 utime_t created
, modified
; // epoch start time
523 int32_t pool_max
; // the largest pool num, ever
527 int num_osd
; // not saved; see calc_num_osds
528 int num_up_osd
; // not saved; see calc_num_osds
529 int num_in_osd
; // not saved; see calc_num_osds
532 std::vector
<uint32_t> osd_state
;
534 mempool::osdmap::map
<int32_t,uint32_t> crush_node_flags
; // crush node -> CEPH_OSD_* flags
535 mempool::osdmap::map
<int32_t,uint32_t> device_class_flags
; // device class -> CEPH_OSD_* flags
537 utime_t last_up_change
, last_in_change
;
539 // These features affect OSDMap[::Incremental] encoding, or the
540 // encoding of some type embedded therein (CrushWrapper, something
541 // from osd_types, etc.).
542 static constexpr uint64_t SIGNIFICANT_FEATURES
=
543 CEPH_FEATUREMASK_PGID64
|
544 CEPH_FEATUREMASK_PGPOOL3
|
545 CEPH_FEATUREMASK_OSDENC
|
546 CEPH_FEATUREMASK_OSDMAP_ENC
|
547 CEPH_FEATUREMASK_OSD_POOLRESEND
|
548 CEPH_FEATUREMASK_NEW_OSDOP_ENCODING
|
549 CEPH_FEATUREMASK_MSG_ADDR2
|
550 CEPH_FEATUREMASK_CRUSH_TUNABLES5
|
551 CEPH_FEATUREMASK_CRUSH_CHOOSE_ARGS
|
552 CEPH_FEATUREMASK_SERVER_LUMINOUS
|
553 CEPH_FEATUREMASK_SERVER_MIMIC
|
554 CEPH_FEATUREMASK_SERVER_NAUTILUS
|
555 CEPH_FEATUREMASK_SERVER_OCTOPUS
;
558 mempool::osdmap::vector
<std::shared_ptr
<entity_addrvec_t
> > client_addrs
;
559 mempool::osdmap::vector
<std::shared_ptr
<entity_addrvec_t
> > cluster_addrs
;
560 mempool::osdmap::vector
<std::shared_ptr
<entity_addrvec_t
> > hb_back_addrs
;
561 mempool::osdmap::vector
<std::shared_ptr
<entity_addrvec_t
> > hb_front_addrs
;
563 std::shared_ptr
<addrs_s
> osd_addrs
;
565 entity_addrvec_t _blank_addrvec
;
567 mempool::osdmap::vector
<__u32
> osd_weight
; // 16.16 fixed point, 0x10000 = "in", 0 = "out"
568 mempool::osdmap::vector
<osd_info_t
> osd_info
;
569 std::shared_ptr
<PGTempMap
> pg_temp
; // temp pg mapping (e.g. while we rebuild)
570 std::shared_ptr
< mempool::osdmap::map
<pg_t
,int32_t > > primary_temp
; // temp primary mapping (e.g. while we rebuild)
571 std::shared_ptr
< mempool::osdmap::vector
<__u32
> > osd_primary_affinity
; ///< 16.16 fixed point, 0x10000 = baseline
573 // remap (post-CRUSH, pre-up)
574 mempool::osdmap::map
<pg_t
,mempool::osdmap::vector
<int32_t>> pg_upmap
; ///< remap pg
575 mempool::osdmap::map
<pg_t
,mempool::osdmap::vector
<std::pair
<int32_t,int32_t>>> pg_upmap_items
; ///< remap osds in up set
577 mempool::osdmap::map
<int64_t,pg_pool_t
> pools
;
578 mempool::osdmap::map
<int64_t,std::string
> pool_name
;
579 mempool::osdmap::map
<std::string
, std::map
<std::string
,std::string
>> erasure_code_profiles
;
580 mempool::osdmap::map
<std::string
,int64_t, std::less
<>> name_pool
;
582 std::shared_ptr
< mempool::osdmap::vector
<uuid_d
> > osd_uuid
;
583 mempool::osdmap::vector
<osd_xinfo_t
> osd_xinfo
;
585 mempool::osdmap::unordered_map
<entity_addr_t
,utime_t
> blocklist
;
587 /// queue of snaps to remove
588 mempool::osdmap::map
<int64_t, snap_interval_set_t
> removed_snaps_queue
;
590 /// removed_snaps additions this epoch
591 mempool::osdmap::map
<int64_t, snap_interval_set_t
> new_removed_snaps
;
593 /// removed_snaps removals this epoch
594 mempool::osdmap::map
<int64_t, snap_interval_set_t
> new_purged_snaps
;
596 epoch_t cluster_snapshot_epoch
;
597 std::string cluster_snapshot
;
598 bool new_blocklist_entries
;
600 float full_ratio
= 0, backfillfull_ratio
= 0, nearfull_ratio
= 0;
602 /// min compat client we want to support
603 ceph_release_t require_min_compat_client
{ceph_release_t::unknown
};
606 /// require osds to run at least this release
607 ceph_release_t require_osd_release
{ceph_release_t::unknown
};
610 mutable uint64_t cached_up_osd_features
;
612 mutable bool crc_defined
;
613 mutable uint32_t crc
;
615 void _calc_up_osd_features();
618 bool have_crc() const { return crc_defined
; }
619 uint32_t get_crc() const { return crc
; }
621 std::shared_ptr
<CrushWrapper
> crush
; // hierarchical map
622 bool stretch_mode_enabled
; // we are in stretch mode, requiring multiple sites
623 uint32_t stretch_bucket_count
; // number of sites we expect to be in
624 uint32_t degraded_stretch_mode
; // 0 if not degraded; else count of up sites
625 uint32_t recovering_stretch_mode
; // 0 if not recovering; else 1
626 int32_t stretch_mode_bucket
; // the bucket type we're stretched across
628 uint32_t crush_version
= 1;
630 friend class OSDMonitor
;
636 num_osd(0), num_up_osd(0), num_in_osd(0),
638 osd_addrs(std::make_shared
<addrs_s
>()),
639 pg_temp(std::make_shared
<PGTempMap
>()),
640 primary_temp(std::make_shared
<mempool::osdmap::map
<pg_t
,int32_t>>()),
641 osd_uuid(std::make_shared
<mempool::osdmap::vector
<uuid_d
>>()),
642 cluster_snapshot_epoch(0),
643 new_blocklist_entries(false),
644 cached_up_osd_features(0),
645 crc_defined(false), crc(0),
646 crush(std::make_shared
<CrushWrapper
>()),
647 stretch_mode_enabled(false), stretch_bucket_count(0),
648 degraded_stretch_mode(0), recovering_stretch_mode(0), stretch_mode_bucket(0) {
652 OSDMap(const OSDMap
& other
) = default;
653 OSDMap
& operator=(const OSDMap
& other
) = default;
656 /// return feature mask subset that is relevant to OSDMap encoding
657 static uint64_t get_significant_features(uint64_t features
) {
658 return SIGNIFICANT_FEATURES
& features
;
661 uint64_t get_encoding_features() const;
663 void deepish_copy_from(const OSDMap
& o
) {
665 primary_temp
.reset(new mempool::osdmap::map
<pg_t
,int32_t>(*o
.primary_temp
));
666 pg_temp
.reset(new PGTempMap(*o
.pg_temp
));
667 osd_uuid
.reset(new mempool::osdmap::vector
<uuid_d
>(*o
.osd_uuid
));
669 if (o
.osd_primary_affinity
)
670 osd_primary_affinity
.reset(new mempool::osdmap::vector
<__u32
>(*o
.osd_primary_affinity
));
672 // NOTE: this still references shared entity_addrvec_t's.
673 osd_addrs
.reset(new addrs_s(*o
.osd_addrs
));
675 // NOTE: we do not copy crush. note that apply_incremental will
676 // allocate a new CrushWrapper, though.
680 const uuid_d
& get_fsid() const { return fsid
; }
681 void set_fsid(uuid_d
& f
) { fsid
= f
; }
683 epoch_t
get_epoch() const { return epoch
; }
684 void inc_epoch() { epoch
++; }
686 void set_epoch(epoch_t e
);
688 uint32_t get_crush_version() const {
689 return crush_version
;
693 const utime_t
& get_created() const { return created
; }
694 const utime_t
& get_modified() const { return modified
; }
696 bool is_blocklisted(const entity_addr_t
& a
) const;
697 bool is_blocklisted(const entity_addrvec_t
& a
) const;
698 void get_blocklist(std::list
<std::pair
<entity_addr_t
,utime_t
> > *bl
) const;
699 void get_blocklist(std::set
<entity_addr_t
> *bl
) const;
701 std::string
get_cluster_snapshot() const {
702 if (cluster_snapshot_epoch
== epoch
)
703 return cluster_snapshot
;
704 return std::string();
707 float get_full_ratio() const {
710 float get_backfillfull_ratio() const {
711 return backfillfull_ratio
;
713 float get_nearfull_ratio() const {
714 return nearfull_ratio
;
716 void get_full_pools(CephContext
*cct
,
717 std::set
<int64_t> *full
,
718 std::set
<int64_t> *backfillfull
,
719 std::set
<int64_t> *nearfull
) const;
720 void get_full_osd_counts(std::set
<int> *full
, std::set
<int> *backfill
,
721 std::set
<int> *nearfull
) const;
724 /***** cluster state *****/
726 int get_max_osd() const { return max_osd
; }
727 void set_max_osd(int m
);
729 unsigned get_num_osds() const {
732 unsigned get_num_up_osds() const {
735 unsigned get_num_in_osds() const {
738 /// recalculate cached values for get_num{,_up,_in}_osds
741 void get_all_osds(std::set
<int32_t>& ls
) const;
742 void get_up_osds(std::set
<int32_t>& ls
) const;
743 void get_out_existing_osds(std::set
<int32_t>& ls
) const;
744 unsigned get_num_pg_temp() const {
745 return pg_temp
->size();
748 int get_flags() const { return flags
; }
749 bool test_flag(int f
) const { return flags
& f
; }
750 void set_flag(int f
) { flags
|= f
; }
751 void clear_flag(int f
) { flags
&= ~f
; }
753 void get_flag_set(std::set
<std::string
> *flagset
) const;
755 static void calc_state_set(int state
, std::set
<std::string
>& st
);
757 int get_state(int o
) const {
758 ceph_assert(o
< max_osd
);
761 int get_state(int o
, std::set
<std::string
>& st
) const {
762 ceph_assert(o
< max_osd
);
763 unsigned t
= osd_state
[o
];
764 calc_state_set(t
, st
);
767 void set_state(int o
, unsigned s
) {
768 ceph_assert(o
< max_osd
);
771 void set_weight(int o
, unsigned w
) {
772 ceph_assert(o
< max_osd
);
775 osd_state
[o
] |= CEPH_OSD_EXISTS
;
777 unsigned get_weight(int o
) const {
778 ceph_assert(o
< max_osd
);
779 return osd_weight
[o
];
781 float get_weightf(int o
) const {
782 return (float)get_weight(o
) / (float)CEPH_OSD_IN
;
784 void adjust_osd_weights(const std::map
<int,double>& weights
, Incremental
& inc
) const;
786 void set_primary_affinity(int o
, int w
) {
787 ceph_assert(o
< max_osd
);
788 if (!osd_primary_affinity
)
789 osd_primary_affinity
.reset(
790 new mempool::osdmap::vector
<__u32
>(
791 max_osd
, CEPH_OSD_DEFAULT_PRIMARY_AFFINITY
));
792 (*osd_primary_affinity
)[o
] = w
;
794 unsigned get_primary_affinity(int o
) const {
795 ceph_assert(o
< max_osd
);
796 if (!osd_primary_affinity
)
797 return CEPH_OSD_DEFAULT_PRIMARY_AFFINITY
;
798 return (*osd_primary_affinity
)[o
];
800 float get_primary_affinityf(int o
) const {
801 return (float)get_primary_affinity(o
) / (float)CEPH_OSD_MAX_PRIMARY_AFFINITY
;
804 bool has_erasure_code_profile(const std::string
&name
) const {
805 auto i
= erasure_code_profiles
.find(name
);
806 return i
!= erasure_code_profiles
.end();
808 int get_erasure_code_profile_default(CephContext
*cct
,
809 std::map
<std::string
,std::string
> &profile_map
,
811 void set_erasure_code_profile(const std::string
&name
,
812 const std::map
<std::string
,std::string
>& profile
) {
813 erasure_code_profiles
[name
] = profile
;
815 const std::map
<std::string
,std::string
> &get_erasure_code_profile(
816 const std::string
&name
) const {
817 static std::map
<std::string
,std::string
> empty
;
818 auto i
= erasure_code_profiles
.find(name
);
819 if (i
== erasure_code_profiles
.end())
824 const mempool::osdmap::map
<std::string
,std::map
<std::string
,std::string
>> &get_erasure_code_profiles() const {
825 return erasure_code_profiles
;
828 bool exists(int osd
) const {
830 return osd
>= 0 && osd
< max_osd
&& (osd_state
[osd
] & CEPH_OSD_EXISTS
);
833 bool is_destroyed(int osd
) const {
834 return exists(osd
) && (osd_state
[osd
] & CEPH_OSD_DESTROYED
);
837 bool is_up(int osd
) const {
838 return exists(osd
) && (osd_state
[osd
] & CEPH_OSD_UP
);
841 bool has_been_up_since(int osd
, epoch_t epoch
) const {
842 return is_up(osd
) && get_up_from(osd
) <= epoch
;
845 bool is_down(int osd
) const {
849 bool is_stop(int osd
) const {
850 return exists(osd
) && is_down(osd
) &&
851 (osd_state
[osd
] & CEPH_OSD_STOP
);
854 bool is_out(int osd
) const {
855 return !exists(osd
) || get_weight(osd
) == CEPH_OSD_OUT
;
858 bool is_in(int osd
) const {
862 bool is_dead(int osd
) const {
864 return false; // unclear if they know they are removed from map
866 return get_xinfo(osd
).dead_epoch
> get_info(osd
).up_from
;
869 unsigned get_osd_crush_node_flags(int osd
) const;
870 unsigned get_crush_node_flags(int id
) const;
871 unsigned get_device_class_flags(int id
) const;
873 bool is_noup_by_osd(int osd
) const {
874 return exists(osd
) && (osd_state
[osd
] & CEPH_OSD_NOUP
);
877 bool is_nodown_by_osd(int osd
) const {
878 return exists(osd
) && (osd_state
[osd
] & CEPH_OSD_NODOWN
);
881 bool is_noin_by_osd(int osd
) const {
882 return exists(osd
) && (osd_state
[osd
] & CEPH_OSD_NOIN
);
885 bool is_noout_by_osd(int osd
) const {
886 return exists(osd
) && (osd_state
[osd
] & CEPH_OSD_NOOUT
);
889 bool is_noup(int osd
) const {
890 if (test_flag(CEPH_OSDMAP_NOUP
)) // global?
892 if (is_noup_by_osd(osd
)) // by osd?
894 if (get_osd_crush_node_flags(osd
) & CEPH_OSD_NOUP
) // by crush-node?
896 if (auto class_id
= crush
->get_item_class_id(osd
); class_id
>= 0 &&
897 get_device_class_flags(class_id
) & CEPH_OSD_NOUP
) // by device-class?
902 bool is_nodown(int osd
) const {
903 if (test_flag(CEPH_OSDMAP_NODOWN
))
905 if (is_nodown_by_osd(osd
))
907 if (get_osd_crush_node_flags(osd
) & CEPH_OSD_NODOWN
)
909 if (auto class_id
= crush
->get_item_class_id(osd
); class_id
>= 0 &&
910 get_device_class_flags(class_id
) & CEPH_OSD_NODOWN
)
915 bool is_noin(int osd
) const {
916 if (test_flag(CEPH_OSDMAP_NOIN
))
918 if (is_noin_by_osd(osd
))
920 if (get_osd_crush_node_flags(osd
) & CEPH_OSD_NOIN
)
922 if (auto class_id
= crush
->get_item_class_id(osd
); class_id
>= 0 &&
923 get_device_class_flags(class_id
) & CEPH_OSD_NOIN
)
928 bool is_noout(int osd
) const {
929 if (test_flag(CEPH_OSDMAP_NOOUT
))
931 if (is_noout_by_osd(osd
))
933 if (get_osd_crush_node_flags(osd
) & CEPH_OSD_NOOUT
)
935 if (auto class_id
= crush
->get_item_class_id(osd
); class_id
>= 0 &&
936 get_device_class_flags(class_id
) & CEPH_OSD_NOOUT
)
942 * check if an entire crush subtree is down
944 bool subtree_is_down(int id
, std::set
<int> *down_cache
) const;
945 bool containing_subtree_is_down(CephContext
*cct
, int osd
, int subtree_type
, std::set
<int> *down_cache
) const;
947 bool subtree_type_is_down(CephContext
*cct
, int id
, int subtree_type
, std::set
<int> *down_in_osds
, std::set
<int> *up_in_osds
,
948 std::set
<int> *subtree_up
, std::unordered_map
<int, std::set
<int> > *subtree_type_down
) const;
950 int identify_osd(const entity_addr_t
& addr
) const;
951 int identify_osd(const uuid_d
& u
) const;
952 int identify_osd_on_all_channels(const entity_addr_t
& addr
) const;
954 bool have_addr(const entity_addr_t
& addr
) const {
955 return identify_osd(addr
) >= 0;
957 int find_osd_on_ip(const entity_addr_t
& ip
) const;
959 const entity_addrvec_t
& get_addrs(int osd
) const {
960 ceph_assert(exists(osd
));
961 return osd_addrs
->client_addrs
[osd
] ?
962 *osd_addrs
->client_addrs
[osd
] : _blank_addrvec
;
964 const entity_addrvec_t
& get_most_recent_addrs(int osd
) const {
965 return get_addrs(osd
);
967 const entity_addrvec_t
&get_cluster_addrs(int osd
) const {
968 ceph_assert(exists(osd
));
969 return osd_addrs
->cluster_addrs
[osd
] ?
970 *osd_addrs
->cluster_addrs
[osd
] : _blank_addrvec
;
972 const entity_addrvec_t
&get_hb_back_addrs(int osd
) const {
973 ceph_assert(exists(osd
));
974 return osd_addrs
->hb_back_addrs
[osd
] ?
975 *osd_addrs
->hb_back_addrs
[osd
] : _blank_addrvec
;
977 const entity_addrvec_t
&get_hb_front_addrs(int osd
) const {
978 ceph_assert(exists(osd
));
979 return osd_addrs
->hb_front_addrs
[osd
] ?
980 *osd_addrs
->hb_front_addrs
[osd
] : _blank_addrvec
;
983 const uuid_d
& get_uuid(int osd
) const {
984 ceph_assert(exists(osd
));
985 return (*osd_uuid
)[osd
];
988 const epoch_t
& get_up_from(int osd
) const {
989 ceph_assert(exists(osd
));
990 return osd_info
[osd
].up_from
;
992 const epoch_t
& get_up_thru(int osd
) const {
993 ceph_assert(exists(osd
));
994 return osd_info
[osd
].up_thru
;
996 const epoch_t
& get_down_at(int osd
) const {
997 ceph_assert(exists(osd
));
998 return osd_info
[osd
].down_at
;
1000 const osd_info_t
& get_info(int osd
) const {
1001 ceph_assert(osd
< max_osd
);
1002 return osd_info
[osd
];
1005 const osd_xinfo_t
& get_xinfo(int osd
) const {
1006 ceph_assert(osd
< max_osd
);
1007 return osd_xinfo
[osd
];
1010 int get_next_up_osd_after(int n
) const {
1011 if (get_max_osd() == 0)
1013 for (int i
= n
+ 1; i
!= n
; ++i
) {
1014 if (i
>= get_max_osd())
1024 int get_previous_up_osd_before(int n
) const {
1025 if (get_max_osd() == 0)
1027 for (int i
= n
- 1; i
!= n
; --i
) {
1029 i
= get_max_osd() - 1;
1039 void get_random_up_osds_by_subtree(int n
, // whoami
1040 std::string
&subtree
,
1041 int limit
, // how many
1043 std::set
<int> *want
) const;
1046 * get feature bits required by the current structure
1048 * @param entity_type [in] what entity type we are asking about
1049 * @param mask [out] std::set of all possible map-related features we could std::set
1050 * @return feature bits used by this map
1052 uint64_t get_features(int entity_type
, uint64_t *mask
) const;
1055 * get oldest *client* version (firefly, hammer, etc.) that can connect given
1056 * the feature bits required (according to get_features()).
1058 ceph_release_t
get_min_compat_client() const;
1061 * gets the required minimum *client* version that can connect to the cluster.
1063 ceph_release_t
get_require_min_compat_client() const;
1066 * get intersection of features supported by up osds
1068 uint64_t get_up_osd_features() const;
1070 void get_upmap_pgs(std::vector
<pg_t
> *upmap_pgs
) const;
1071 bool check_pg_upmaps(
1073 const std::vector
<pg_t
>& to_check
,
1074 std::vector
<pg_t
> *to_cancel
,
1075 std::map
<pg_t
, mempool::osdmap::vector
<std::pair
<int,int>>> *to_remap
) const;
1076 void clean_pg_upmaps(
1078 Incremental
*pending_inc
,
1079 const std::vector
<pg_t
>& to_cancel
,
1080 const std::map
<pg_t
, mempool::osdmap::vector
<std::pair
<int,int>>>& to_remap
) const;
1081 bool clean_pg_upmaps(CephContext
*cct
, Incremental
*pending_inc
) const;
1083 int apply_incremental(const Incremental
&inc
);
1085 /// try to re-use/reference addrs in oldmap from newmap
1086 static void dedup(const OSDMap
*oldmap
, OSDMap
*newmap
);
1088 static void clean_temps(CephContext
*cct
,
1089 const OSDMap
& oldmap
,
1090 const OSDMap
& nextmap
,
1091 Incremental
*pending_inc
);
1093 // serialize, unserialize
1095 void encode_client_old(ceph::buffer::list
& bl
) const;
1096 void encode_classic(ceph::buffer::list
& bl
, uint64_t features
) const;
1097 void decode_classic(ceph::buffer::list::const_iterator
& p
);
1100 void encode(ceph::buffer::list
& bl
, uint64_t features
=CEPH_FEATURES_ALL
) const;
1101 void decode(ceph::buffer::list
& bl
);
1102 void decode(ceph::buffer::list::const_iterator
& bl
);
1105 /**** mapping facilities ****/
1108 const std::string
& name
,
1109 const std::string
& key
,
1110 const std::string
& nspace
,
1112 int object_locator_to_pg(const object_t
& oid
, const object_locator_t
& loc
,
1114 pg_t
object_locator_to_pg(const object_t
& oid
,
1115 const object_locator_t
& loc
) const {
1117 int ret
= object_locator_to_pg(oid
, loc
, pg
);
1118 ceph_assert(ret
== 0);
1123 static object_locator_t
file_to_object_locator(const file_layout_t
& layout
) {
1124 return object_locator_t(layout
.pool_id
, layout
.pool_ns
);
1127 ceph_object_layout
file_to_object_layout(object_t oid
,
1128 file_layout_t
& layout
) const {
1129 return make_object_layout(oid
, layout
.pool_id
, layout
.pool_ns
);
1132 ceph_object_layout
make_object_layout(object_t oid
, int pg_pool
,
1133 std::string nspace
) const;
1135 int get_pg_num(int pg_pool
) const
1137 const pg_pool_t
*pool
= get_pg_pool(pg_pool
);
1138 ceph_assert(NULL
!= pool
);
1139 return pool
->get_pg_num();
1142 bool pg_exists(pg_t pgid
) const {
1143 const pg_pool_t
*p
= get_pg_pool(pgid
.pool());
1144 return p
&& pgid
.ps() < p
->get_pg_num();
1147 int get_pg_pool_min_size(pg_t pgid
) const {
1148 if (!pg_exists(pgid
)) {
1151 const pg_pool_t
*p
= get_pg_pool(pgid
.pool());
1153 return p
->get_min_size();
1156 int get_pg_pool_size(pg_t pgid
) const {
1157 if (!pg_exists(pgid
)) {
1160 const pg_pool_t
*p
= get_pg_pool(pgid
.pool());
1162 return p
->get_size();
1165 int get_pg_pool_crush_rule(pg_t pgid
) const {
1166 if (!pg_exists(pgid
)) {
1169 const pg_pool_t
*p
= get_pg_pool(pgid
.pool());
1171 return p
->get_crush_rule();
1175 /// pg -> (raw osd std::list)
1176 void _pg_to_raw_osds(
1177 const pg_pool_t
& pool
, pg_t pg
,
1178 std::vector
<int> *osds
,
1180 int _pick_primary(const std::vector
<int>& osds
) const;
1181 void _remove_nonexistent_osds(const pg_pool_t
& pool
, std::vector
<int>& osds
) const;
1183 void _apply_primary_affinity(ps_t seed
, const pg_pool_t
& pool
,
1184 std::vector
<int> *osds
, int *primary
) const;
1186 /// apply pg_upmap[_items] mappings
1187 void _apply_upmap(const pg_pool_t
& pi
, pg_t pg
, std::vector
<int> *raw
) const;
1189 /// pg -> (up osd std::list)
1190 void _raw_to_up_osds(const pg_pool_t
& pool
, const std::vector
<int>& raw
,
1191 std::vector
<int> *up
) const;
1195 * Get the pg and primary temp, if they are specified.
1196 * @param temp_pg [out] Will be empty or contain the temp PG mapping on return
1197 * @param temp_primary [out] Will be the value in primary_temp, or a value derived
1198 * from the pg_temp (if specified), or -1 if you should use the calculated (up_)primary.
1200 void _get_temp_osds(const pg_pool_t
& pool
, pg_t pg
,
1201 std::vector
<int> *temp_pg
, int *temp_primary
) const;
1204 * map to up and acting. Fills in whatever fields are non-NULL.
1206 void _pg_to_up_acting_osds(const pg_t
& pg
, std::vector
<int> *up
, int *up_primary
,
1207 std::vector
<int> *acting
, int *acting_primary
,
1208 bool raw_pg_to_pg
= true) const;
1212 * This is suitable only for looking at raw CRUSH outputs. It skips
1213 * applying the temp and up checks and should not be used
1214 * by anybody for data mapping purposes.
1215 * raw and primary must be non-NULL
1217 void pg_to_raw_osds(pg_t pg
, std::vector
<int> *raw
, int *primary
) const;
1218 void pg_to_raw_upmap(pg_t pg
, std::vector
<int> *raw
,
1219 std::vector
<int> *raw_upmap
) const;
1220 /// map a pg to its acting set. @return acting set size
1221 void pg_to_acting_osds(const pg_t
& pg
, std::vector
<int> *acting
,
1222 int *acting_primary
) const {
1223 _pg_to_up_acting_osds(pg
, NULL
, NULL
, acting
, acting_primary
);
1225 void pg_to_acting_osds(pg_t pg
, std::vector
<int>& acting
) const {
1226 return pg_to_acting_osds(pg
, &acting
, NULL
);
1229 * This does not apply temp overrides and should not be used
1230 * by anybody for data mapping purposes. Specify both pointers.
1232 void pg_to_raw_up(pg_t pg
, std::vector
<int> *up
, int *primary
) const;
1234 * map a pg to its acting set as well as its up set. You must use
1235 * the acting set for data mapping purposes, but some users will
1236 * also find the up set useful for things like deciding what to
1238 * Each of these pointers must be non-NULL.
1240 void pg_to_up_acting_osds(pg_t pg
, std::vector
<int> *up
, int *up_primary
,
1241 std::vector
<int> *acting
, int *acting_primary
) const {
1242 _pg_to_up_acting_osds(pg
, up
, up_primary
, acting
, acting_primary
);
1244 void pg_to_up_acting_osds(pg_t pg
, std::vector
<int>& up
, std::vector
<int>& acting
) const {
1245 int up_primary
, acting_primary
;
1246 pg_to_up_acting_osds(pg
, &up
, &up_primary
, &acting
, &acting_primary
);
1248 bool pg_is_ec(pg_t pg
) const {
1249 auto i
= pools
.find(pg
.pool());
1250 ceph_assert(i
!= pools
.end());
1251 return i
->second
.is_erasure();
1253 bool get_primary_shard(const pg_t
& pgid
, spg_t
*out
) const {
1254 auto i
= get_pools().find(pgid
.pool());
1255 if (i
== get_pools().end()) {
1258 if (!i
->second
.is_erasure()) {
1263 std::vector
<int> acting
;
1264 pg_to_acting_osds(pgid
, &acting
, &primary
);
1265 for (uint8_t i
= 0; i
< acting
.size(); ++i
) {
1266 if (acting
[i
] == primary
) {
1267 *out
= spg_t(pgid
, shard_id_t(i
));
1273 bool get_primary_shard(const pg_t
& pgid
, int *primary
, spg_t
*out
) const {
1274 auto i
= get_pools().find(pgid
.pool());
1275 if (i
== get_pools().end()) {
1278 std::vector
<int> acting
;
1279 pg_to_acting_osds(pgid
, &acting
, primary
);
1280 if (i
->second
.is_erasure()) {
1281 for (uint8_t i
= 0; i
< acting
.size(); ++i
) {
1282 if (acting
[i
] == *primary
) {
1283 *out
= spg_t(pgid
, shard_id_t(i
));
1294 bool in_removed_snaps_queue(int64_t pool
, snapid_t snap
) const {
1295 auto p
= removed_snaps_queue
.find(pool
);
1296 if (p
== removed_snaps_queue
.end()) {
1299 return p
->second
.contains(snap
);
1302 const mempool::osdmap::map
<int64_t,snap_interval_set_t
>&
1303 get_removed_snaps_queue() const {
1304 return removed_snaps_queue
;
1306 const mempool::osdmap::map
<int64_t,snap_interval_set_t
>&
1307 get_new_removed_snaps() const {
1308 return new_removed_snaps
;
1310 const mempool::osdmap::map
<int64_t,snap_interval_set_t
>&
1311 get_new_purged_snaps() const {
1312 return new_purged_snaps
;
1315 int64_t lookup_pg_pool_name(std::string_view name
) const {
1316 auto p
= name_pool
.find(name
);
1317 if (p
== name_pool
.end())
1322 int64_t get_pool_max() const {
1325 const mempool::osdmap::map
<int64_t,pg_pool_t
>& get_pools() const {
1328 mempool::osdmap::map
<int64_t,pg_pool_t
>& get_pools() {
1331 void get_pool_ids_by_rule(int rule_id
, std::set
<int64_t> *pool_ids
) const {
1332 ceph_assert(pool_ids
);
1333 for (auto &p
: pools
) {
1334 if (p
.second
.get_crush_rule() == rule_id
) {
1335 pool_ids
->insert(p
.first
);
1339 void get_pool_ids_by_osd(CephContext
*cct
,
1341 std::set
<int64_t> *pool_ids
) const;
1342 const std::string
& get_pool_name(int64_t p
) const {
1343 auto i
= pool_name
.find(p
);
1344 ceph_assert(i
!= pool_name
.end());
1347 const mempool::osdmap::map
<int64_t,std::string
>& get_pool_names() const {
1350 bool have_pg_pool(int64_t p
) const {
1351 return pools
.count(p
);
1353 const pg_pool_t
* get_pg_pool(int64_t p
) const {
1354 auto i
= pools
.find(p
);
1355 if (i
!= pools
.end())
1359 unsigned get_pg_size(pg_t pg
) const {
1360 auto p
= pools
.find(pg
.pool());
1361 ceph_assert(p
!= pools
.end());
1362 return p
->second
.get_size();
1364 int get_pg_type(pg_t pg
) const {
1365 auto p
= pools
.find(pg
.pool());
1366 ceph_assert(p
!= pools
.end());
1367 return p
->second
.get_type();
1369 int get_pool_crush_rule(int64_t pool_id
) const {
1370 auto pool
= get_pg_pool(pool_id
);
1373 return pool
->get_crush_rule();
1377 pg_t
raw_pg_to_pg(pg_t pg
) const {
1378 auto p
= pools
.find(pg
.pool());
1379 ceph_assert(p
!= pools
.end());
1380 return p
->second
.raw_pg_to_pg(pg
);
1383 // pg -> acting primary osd
1384 int get_pg_acting_primary(pg_t pg
) const {
1386 _pg_to_up_acting_osds(pg
, nullptr, nullptr, nullptr, &primary
);
1391 * check whether an spg_t maps to a particular osd
1393 bool is_up_acting_osd_shard(spg_t pg
, int osd
) const {
1394 std::vector
<int> up
, acting
;
1395 _pg_to_up_acting_osds(pg
.pgid
, &up
, NULL
, &acting
, NULL
, false);
1396 if (calc_pg_role(pg_shard_t(osd
, pg
.shard
), acting
) >= 0 ||
1397 calc_pg_role(pg_shard_t(osd
, pg
.shard
), up
) >= 0) {
1404 static int calc_pg_role_broken(int osd
, const std::vector
<int>& acting
, int nrep
=0);
1405 static int calc_pg_role(pg_shard_t who
, const std::vector
<int>& acting
);
1406 static bool primary_changed_broken(
1408 const std::vector
<int> &oldacting
,
1410 const std::vector
<int> &newacting
);
1412 /* rank is -1 (stray), 0 (primary), 1,2,3,... (replica) */
1413 int get_pg_acting_role(spg_t pg
, int osd
) const {
1414 std::vector
<int> group
;
1415 pg_to_acting_osds(pg
.pgid
, group
);
1416 return calc_pg_role(pg_shard_t(osd
, pg
.shard
), group
);
1421 pg_t pg
, ///< pg to potentially remap
1422 const std::set
<int>& overfull
, ///< osds we'd want to evacuate
1423 const std::vector
<int>& underfull
, ///< osds to move to, in order of preference
1424 const std::vector
<int>& more_underfull
, ///< less full osds to move to, in order of preference
1425 std::vector
<int> *orig
,
1426 std::vector
<int> *out
); ///< resulting alternative mapping
1430 uint32_t max_deviation
, ///< max deviation from target (value >= 1)
1431 int max_iterations
, ///< max iterations to run
1432 const std::set
<int64_t>& pools
, ///< [optional] restrict to pool
1433 Incremental
*pending_inc
1436 int get_osds_by_bucket_name(const std::string
&name
, std::set
<int> *osds
) const;
1438 bool have_pg_upmaps(pg_t pg
) const {
1439 return pg_upmap
.count(pg
) ||
1440 pg_upmap_items
.count(pg
);
1443 bool check_full(const std::set
<pg_shard_t
> &missing_on
) const {
1444 for (auto shard
: missing_on
) {
1445 if (get_state(shard
.osd
) & CEPH_OSD_FULL
)
1452 * handy helpers to build simple maps...
1455 * Build an OSD map suitable for basic usage. If **num_osd** is >= 0
1456 * it will be initialized with the specified number of OSDs in a
1457 * single host. If **num_osd** is < 0 the layout of the OSD map will
1458 * be built by reading the content of the configuration file.
1460 * @param cct [in] in core ceph context
1461 * @param e [in] initial epoch
1462 * @param fsid [in] id of the cluster
1463 * @param num_osd [in] number of OSDs if >= 0 or read from conf if < 0
1464 * @return **0** on success, negative errno on error.
1467 int build_simple_optioned(CephContext
*cct
, epoch_t e
, uuid_d
&fsid
,
1468 int num_osd
, int pg_bits
, int pgp_bits
,
1471 int build_simple(CephContext
*cct
, epoch_t e
, uuid_d
&fsid
,
1473 return build_simple_optioned(cct
, e
, fsid
, num_osd
, 0, 0, false);
1475 int build_simple_with_pool(CephContext
*cct
, epoch_t e
, uuid_d
&fsid
,
1476 int num_osd
, int pg_bits
, int pgp_bits
) {
1477 return build_simple_optioned(cct
, e
, fsid
, num_osd
,
1478 pg_bits
, pgp_bits
, true);
1480 static int _build_crush_types(CrushWrapper
& crush
);
1481 static int build_simple_crush_map(CephContext
*cct
, CrushWrapper
& crush
,
1482 int num_osd
, std::ostream
*ss
);
1483 static int build_simple_crush_map_from_conf(CephContext
*cct
,
1484 CrushWrapper
& crush
,
1486 static int build_simple_crush_rules(
1487 CephContext
*cct
, CrushWrapper
& crush
,
1488 const std::string
& root
,
1491 bool crush_rule_in_use(int rule_id
) const;
1493 int validate_crush_rules(CrushWrapper
*crush
, std::ostream
*ss
) const;
1497 primary_temp
->clear();
1501 void print_osd_line(int cur
, std::ostream
*out
, ceph::Formatter
*f
) const;
1503 void print(std::ostream
& out
) const;
1504 void print_osd(int id
, std::ostream
& out
) const;
1505 void print_osds(std::ostream
& out
) const;
1506 void print_pools(std::ostream
& out
) const;
1507 void print_summary(ceph::Formatter
*f
, std::ostream
& out
,
1508 const std::string
& prefix
, bool extra
=false) const;
1509 void print_oneline_summary(std::ostream
& out
) const;
1512 DUMP_IN
= 1, // only 'in' osds
1513 DUMP_OUT
= 2, // only 'out' osds
1514 DUMP_UP
= 4, // only 'up' osds
1515 DUMP_DOWN
= 8, // only 'down' osds
1516 DUMP_DESTROYED
= 16, // only 'destroyed' osds
1518 void print_tree(ceph::Formatter
*f
, std::ostream
*out
,
1519 unsigned dump_flags
=0, std::string bucket
="") const;
1521 int summarize_mapping_stats(
1523 const std::set
<int64_t> *pools
,
1525 ceph::Formatter
*f
) const;
1527 std::string
get_flag_string() const;
1528 static std::string
get_flag_string(unsigned flags
);
1529 static void dump_erasure_code_profiles(
1530 const mempool::osdmap::map
<std::string
,std::map
<std::string
,std::string
> > &profiles
,
1531 ceph::Formatter
*f
);
1532 void dump(ceph::Formatter
*f
) const;
1533 void dump_osd(int id
, ceph::Formatter
*f
) const;
1534 void dump_osds(ceph::Formatter
*f
) const;
1535 static void generate_test_instances(std::list
<OSDMap
*>& o
);
1536 bool check_new_blocklist_entries() const { return new_blocklist_entries
; }
1538 void check_health(CephContext
*cct
, health_check_map_t
*checks
) const;
1540 int parse_osd_id_list(const std::vector
<std::string
>& ls
,
1542 std::ostream
*ss
) const;
1544 float pool_raw_used_rate(int64_t poolid
) const;
1547 WRITE_CLASS_ENCODER_FEATURES(OSDMap
)
1548 WRITE_CLASS_ENCODER_FEATURES(OSDMap::Incremental
)
1551 using OSDMapRef
= boost::local_shared_ptr
<const OSDMap
>;
1553 using OSDMapRef
= std::shared_ptr
<const OSDMap
>;
1557 inline std::ostream
& operator<<(std::ostream
& out
, const OSDMap
& m
) {
1558 m
.print_oneline_summary(out
);
1564 void print_osd_utilization(const OSDMap
& osdmap
,
1569 const std::string
& filter
);