Our test admin has been asking for this for the past few years:-)
Besides, this is also useful for operating on large Ceph clusters with
mutliple storage pools possibly spanning over all osds.
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
Due to pipe, sdiff return code was ignored. Also the compare functions
spent most of time in xxd, which was normally unnecessary.
Signed-off-by: Mykola Golub <mgolub@suse.com>
to build helloworld of librados, we need to install the build
dependencies.
Fixes: https://tracker.ceph.com/issues/40155
Signed-off-by: Kefu Chai <kchai@redhat.com>
It does not make much sense to add this kind of restrictions
as long as user is aware of what is going on.
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
The "unmap" request is asynchronous, so wait for a short amount
of time for the "rbd-nbd" daemon process to exit.
Fixes: http://tracker.ceph.com/issues/39598
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
We currently do not expose perf counters in the restful module but this
is a useful piece of information to be exposed by the restful module.
Since we already have a function to gather all the information about the
perf counters, we can make this easily accessible through the restful
API.
This also supports filtering by daemons through regexps.
Signed-off-by: Boris Ranto <branto@redhat.com>
Use OSD_POOL_PRIORITY_MAX and OSD_POOL_PRIORITY_MIN constants
Scale legacy priorities if exceeds maximum
Signed-off-by: David Zafman <dzafman@redhat.com>
This was used for API backwards compatibility testing, but now that
the C++ API will not remain stable, it serves no purpose.
Fixes: http://tracker.ceph.com/issues/39072
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
With discard_granularity set to alloc_size, we no longer get object
size alignment from blk_bio_discard_split().
This assumption is pretty deeply ingrained in krbd_data_pool.sh, so
make it explicit. For krbd_fallocate.sh, just fix the expectation.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
The test case is issuing discards that span two objects: the tail of
the first object is truncated, the head of the second object is zeroed.
These discards aren't serial, so there is a race:
discard i ~ i + 1: truncate i, zero i + 1
discard i + 1 ~ i + 2: truncate i + 1, zero i + 2
can be executed as
truncate i + 1, zero i + 2, truncate i, zero i + 1
For object i + 1, the sequence ends up being truncate tail, then zero
head. This zero op is munged to truncate on the OSD, resulting in size
0 instead of OBJECT_SIZE / 2.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
This way it can be used to fast cancel/undo the last command.
Also make the tip message a litter bit nicer..
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
There is a lot of good stuff going on here, but nobody is investing in xio
and it is not expected to be the path forward for RDMA. If that ever
changes, we can resurrect the code. Until then, let's clean up the tree
and reduce friction for changes going forward.
Signed-off-by: Sage Weil <sage@redhat.com>
* refs/pull/26894/head:
qa/standalone/erasure-code/test-erasure-code: adjust test to avoid m=0
erasure-code: ensure m >= 1
mon/OSDMonitor: set ec min_size to k + min(1, m - 1)
Reviewed-by: Kefu Chai <kchai@redhat.com>
Reviewed-by: Neha Ojha <nojha@redhat.com>
The crush/builder.c crush_add_bucket method resizes the max_buckets array
but a power of 2 when it has to expand, but the code in CrushWrapper was
assuming that if the array grew the pos for the new bucket would be the
last position in the new array. This led to a situation where the
crush_choose_arg_map args array size didn't match max_buckets, and
eventually caused a crash.
Fixes: http://tracker.ceph.com/issues/38664
Signed-off-by: Sage Weil <sage@redhat.com>
For large clusters, we use device classes to isolate storage pools.
The existing 'osd df' output turns out to be too nosiy, say, if
you care about only single storage pool with osds possibly spanning over
all hosts.
With this change you are now being able to do 'osd df' by class (or by pool,
if you simply use classes to separate different pools), or by a specified
crush bucket name you are currently interested in, which is much more
convenient.
Some examples:
```
$ bin/ceph osd df tree
ID CLASS WEIGHT REWEIGHT SIZE RAW USE DATA OMAP META AVAIL %USE VAR PGS STATUS TYPE NAME
-1 0.05878 - 60 GiB 6.4 GiB 23 MiB 0 B 6 GiB 54 GiB 10.60 1.00 - root default
-3 0.02939 - 30 GiB 3.2 GiB 12 MiB 0 B 3 GiB 27 GiB 10.60 1.00 - host ceph11
3 aaa 0.00980 1.00000 10 GiB 1.1 GiB 3.9 MiB 0 B 1 GiB 9.0 GiB 10.60 1.00 56 up osd.3
4 bbb 0.00980 1.00000 10 GiB 1.1 GiB 3.9 MiB 0 B 1 GiB 9.0 GiB 10.60 1.00 58 up osd.4
5 ccc 0.00980 1.00000 10 GiB 1.1 GiB 3.9 MiB 0 B 1 GiB 9.0 GiB 10.60 1.00 60 up osd.5
-5 0.02939 - 30 GiB 3.2 GiB 12 MiB 0 B 3 GiB 27 GiB 10.60 1.00 - host ceph12
0 aaa 0.00980 1.00000 10 GiB 1.1 GiB 3.9 MiB 0 B 1 GiB 9.0 GiB 10.60 1.00 50 up osd.0
1 bbb 0.00980 1.00000 10 GiB 1.1 GiB 3.9 MiB 0 B 1 GiB 9.0 GiB 10.60 1.00 61 up osd.1
2 ccc 0.00980 1.00000 10 GiB 1.1 GiB 3.9 MiB 0 B 1 GiB 9.0 GiB 10.60 1.00 51 up osd.2
TOTAL 60 GiB 6.4 GiB 23 MiB 0 B 6 GiB 54 GiB 10.60
MIN/MAX VAR: 1.00/1.00 STDDEV: 0
$ bin/ceph osd df tree class aaa
ID CLASS WEIGHT REWEIGHT SIZE RAW USE DATA OMAP META AVAIL %USE VAR PGS STATUS TYPE NAME
-1 0.05878 - 20 GiB 2.1 GiB 7.8 MiB 0 B 2 GiB 18 GiB 10.60 1.00 - root default
-3 0.02939 - 10 GiB 1.1 GiB 3.9 MiB 0 B 1 GiB 9.0 GiB 10.60 1.00 - host ceph11
3 aaa 0.00980 1.00000 10 GiB 1.1 GiB 3.9 MiB 0 B 1 GiB 9.0 GiB 10.60 1.00 56 up osd.3
-5 0.02939 - 10 GiB 1.1 GiB 3.9 MiB 0 B 1 GiB 9.0 GiB 10.60 1.00 - host ceph12
0 aaa 0.00980 1.00000 10 GiB 1.1 GiB 3.9 MiB 0 B 1 GiB 9.0 GiB 10.60 1.00 50 up osd.0
TOTAL 20 GiB 2.1 GiB 7.8 MiB 0 B 2 GiB 18 GiB 10.60
MIN/MAX VAR: 1.00/1.00 STDDEV: 0
$ bin/ceph osd df tree name ceph11
ID CLASS WEIGHT REWEIGHT SIZE RAW USE DATA OMAP META AVAIL %USE VAR PGS STATUS TYPE NAME
-3 0.02939 - 30 GiB 3.2 GiB 12 MiB 0 B 3 GiB 27 GiB 10.60 1.00 - host ceph11
3 aaa 0.00980 1.00000 10 GiB 1.1 GiB 3.9 MiB 0 B 1 GiB 9.0 GiB 10.60 1.00 56 up osd.3
4 bbb 0.00980 1.00000 10 GiB 1.1 GiB 3.9 MiB 0 B 1 GiB 9.0 GiB 10.60 1.00 58 up osd.4
5 ccc 0.00980 1.00000 10 GiB 1.1 GiB 3.9 MiB 0 B 1 GiB 9.0 GiB 10.60 1.00 60 up osd.5
TOTAL 30 GiB 3.2 GiB 12 MiB 0 B 3 GiB 27 GiB 10.60
MIN/MAX VAR: 1.00/1.00 STDDEV: 0
```
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
These have bit-rotted and no longer work. No cycles from interested parties
available to fix.
Fixes: https://tracker.ceph.com/issues/38487
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
This better mimics the behavior of teuthology and tests rbd-mirror
daemon's ability to handle a pool deletion.
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
For those with multiple storage pools sharing the same devices,
I think it would make much more sense to offer per-pool
commands to bring pools with high priority, e.g., because they
are hosting data of more importance than others, back to normal
quickly.
Fixes: http://tracker.ceph.com/issues/38456
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
* refs/pull/26237/head:
Revert "qa: update test_envlibrados_for_rocksdb.sh for libradospp split"
doc/librados: explicitly mention that the C++ API is not stable
ceph.spec: force use of upgrade devtoolset-gcc under RHEL 7
librados: add symbol versioning to the C++ API
librados: add symbol versioning to the C API
librados: revert librados3/libradoscc back to librados2
Reviewed-by: Kefu Chai <kchai@redhat.com>
The test extracts the mon addresses from the monmap, but with the
recent v2 format change it extracted an invalid address.
Fixes: http://tracker.ceph.com/issues/38385
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
* refs/pull/24805/head:
qa/suite: add dedup test
src/tools: fix compile error (master version issue)
src/tools: add stats (fixed objects,total objects)
src/tools: make room for cdc
src/tools: make enhacned stats and interface class
src/tools: set timelimit and add signal handler to check progress
src/tools: use the slice thing and make parallel (chunk_scrub)
src/test: add max-thread test in test_dedup_tool.sh
src/tools: use the slice thing and make parallel
src/test: add chunk-scrub test in test_dedup_tool.sh
src/tools: add chunk-scrub op in dedup tool
src/cls/cas: add has_chunk op
src/test: add test_dedup_tool.sh
src/tools: initial works for dedup tool
Reviewed-by: Sage Weil <sage@redhat.com>
The pool and namespace can now be specified as in a
<pool-name>[/<namespace-name>] format as positional
arguments.
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
* refs/pull/25009/head:
librbd: stringify locker name with get_legacy_str()
osdc/Objecter: fix list_watchers addr rendering to match legacy
test/crimson: disable unittest_seastar_messenger test
msg/msg_types: encode entity_addr_t TYPE_ANY as TYPE_LEGACY for pre-nautilus
client: make blacklist detection handle TYPE_ANY entries
mon/OSDMonitor: maintain compat output for 'blacklist ls'
client: maintain compat for {inst,addr}_str in status dump
qa/tasks/ceph_manager: compare osd flush seq #'s as ints
qa/suites/fs: make use of simple.yaml where appropriate
qa/msgr: move msgr factet into generic re-usable dir
crimson: fix monmap build for seastar
doc/start/ceph.conf: trim the sample ceph.conf file
doc/rados/operations: only describe --public-{addr,network} method for adding mons
PendingReleaseNotes: deprecate 'mon addr'
doc: fix some 'mon addr' references
doc/rados/configuration: fix some 'mon addr' references
doc/rados/configuration/network-config-ref: revise network docs somewhat
doc/rados/configuration/network-config-ref: remove totally obsolete section
qa/suites/rados: replace mon_seesaw.py task with a small bash script
qa/suites/fs/upgrade: don't bind to v2 addrs
qa/tasks/mon_thrash: avoid 'mon addr' in mon section
mon/MonClient: disable ms_bind_msgr2 if NAUTILUS feature not set
osd/OSDMap: maintain compat addr fields
msg/msg_types: add get_legacy_str()
mds/MDSMap.h: maintain compat addr field
mon/MgrMap: maintain compat active_addr field
mon/MonClient: reconnect to mon if it's addrvec appears to have changed
qa/tasks/ceph.conf.template: increase mon_mgr_mkfs_grace
msg/async/ProtocolV2: fill in IP for all peer_addrs
msg/async: print all addrs on debug lines
mon/MonMap: no noname- mon name prefix when for_mkfs
ceph-monstore-tool: print initial monmap
msg/async/ProtocolV2: advertise ourselves as a v2 addr when using v2 protocol
msg/async: assert existing protocol matches current protocol
msg/async: add missing modelines
mon/MonMap: add missing modeline
vstart.sh: put mon addrs in mon_host, not 'mon addr'
msg/async: better debug around conn map lookups and updates
mon/MonClient: dump initial monmap at debug level 10
qa/standalone/osd/osd-fast-mark-down: use v1 addr w/ simplemessenger
qa/tasks/ceph: set initial monmap features with using addrvec addrs
monmaptool: add --enable-all-features option
qa/tasks/ceph: only use monmaptool --addv if addr has [,:v]
qa/tasks/ceph_manager: make get_mon_status use mon addr
qa/tasks/ceph: keep mon addrs in ctx namespace
mon/OSDMonitor: log all osd addrs on boot
msg/simple: behave when v2 and v1 addrs are present at target
mon/MonClient: warn if global_id changes
msg/Connection: add warning/note on get_peer_global_id
mds/MDSDaemon: clean up handle_mds_map debug output a bit
qa/suites/rados/upgrade: debug mds
mds/MDSRank: improve is_stale_message to handle addrvecs
msg/async: make loopback detect when sending to one of our many addrs
qa/suites/rados/upgrade: no aggressive pg num changes
mon/OSDMonitor: require nautilus mons for require_osd_release=nautilus
mon/OSDMonitor: require mimic mons for require_osd_release=mimic
qa/suites/rados/thrash-old-clients: use legacy addr syntax in ceph.conf
msg/async: preserve peer features when replacing a connection
qa/tasks/ceph.py: move methods from teuthology.git into ceph.py directly; support mon bind * options
mon/MonMap: adjust build_initial behavior for mkfs vs probe
mon/MonMap: improve ambiguous addr behavior
qa/suites/rados/upgrade: spread mons a bit
qa/rados/thrash-old-clients: keep mons on separate hosts
qa/standalone/mon/misc.sh: tweak test to be more robust
qa/tasks/mon_seesaw: expect v1/v2 prefix in addr
osd/OSDMap: fix is_blacklisted() check to assume type ANY
mon/OSDMonitor: use ANY addr type for blacklisting
mon/msg_types: TYPE_V1ORV2 -> TYPE_ANY
qa/workunits/cephtool: fix blacklist test
qa/suites/upgrade: install old version with only v1 addrs
common/options: by default, bind to both msgr v1 and v2 addresses
vstart.sh: add --msgr1, --msgr2, --msgr21 options
msg/async/ProtocolV2: be flexible with server identity check
msg/msg_types: fix entity_addrvec_t::parse() with null end arg
qa/suites/rados/basic/msgr: no msgr2 addrs in initial monmaps
qa/tasks/ceph: add 'mon_bind_addrvec' and 'mon_bind_msgr2' options
monmaptool: add --addv argument to pass in addrvec directly
qa/suites/rados/basic/msgr: do not use msgr2 with simplemessenger
qa/suites/rados/basic/msgr: async is not experimental
messages/MOSDBoot: fix compat with pre-nautilus
mon/MonMap: allow v1 or v2 to be explicitly specified along with part
msg/msg_types: allow parsing of IPs without assuming v1 vs v2
msg/msg_types: default parse to v2 addrs
msg: standarize on v1: and v2: prefixes for *all* entity_addr_t's
vstart.sh: use msgr2 by default
mon/MonMap: remove get_addr() methods
ceph-mon: adjust startup/bind/join sequence to use addrs
mon: use MonMap::get_addrs() (instead of get_addr())
mon/MonClient: change pending_cons to addrvec-based map
mon/MonMap: fix set_addr() caller, kill wrapper
mon/MonMap: remove addr-based add()
monmaptool: fix --add to do either legacy or msgr2+legacy
monmaptool: clean up iterator use a bit
mon/MonMap: handle ambiguous mon addrs by trying both legacy and msgr
mon/MonMap: take addrvec for set_initial_members
mon/MonMap: use addrvecs for test instances
mon: pass addrvec via MMonJoin
mon/MonmapMonitor: fix 'mon add' to populate addrvec
mon/MonMap: addr -> addrvec
msg/async/ProtocolV2: only update socket_addr if we learned our addr
osd: go active even if mon only accepted our v1 addr
test/msgr: add test for msgr2 protocol
msg/async/ProtocolV2: share socket_addr and all addrs during handshake
msg/async: print socket_addr for the connection
msg/async: msgr2 protocol placeholder
msg/async: move ProtocolV1 class to its own source file
msg/async: keep listen addr in ServerSocket, pass to new connections
msg/async/AsyncMessenger: fix set_addr_unknowns
Reviewed-by: Josh Durgin <jdurgin@redhat.com>
This avoids a huge pg merge from 100s to 4, which takes a long time and
makes the teuthology scrub cleanup time out.
Signed-off-by: Sage Weil <sage@redhat.com>
This isn't really relevant or useful now that the mgr is throttling the
actual pg_num adjustment based on pg_num_target, % misplaced, etc.
Signed-off-by: Sage Weil <sage@redhat.com>
* refs/pull/25190/head:
mgr/prometheus: adjust to new 'df' fields
mon/Monitor: fix newline between df section
doc: update docs for new ceph df output
mon/PGMap: break down RAW usage by device class
mon/PGMap: tweak df headers
mon/PGMap: GLOBAL -> RAW STORAGE in 'df' output
mon/PGMap: dump_fs_stats -> dump_cluster_stats
Reviewed-by: Kefu Chai <kchai@redhat.com>
before this change, we assume that the variable set if rados::radospp is
found will be radospp_FOUND, but this is a feature cmake 3, see
https://cmake.org/cmake/help/v3.3/module/FindPackageHandleStandardArgs.html
while the cmake shipped by centos is cmake 2.8.12, where the variable
name will be <UPPERCASED_NAME>_FOUND, see
https://cmake.org/cmake/help/v2.8.12/cmake.html#module:FindPackageHandleStandardArgs
in the test of test_envlibrados_for_rocksdb.sh, we are using cmake not
the cmake3 offered by EPEL7, so RADOSPP_FOUND will be set instead. that's why
executable env_librados_test will fail to link against rados::radospp.
as rados::radospp won't be defined if radospp_FOUND is not defined/set.
after this change, the 2nd mode of FIND_PACKAGE_HANDLE_STANDARD_ARGS()
is used instead to ensure that radospp_FOUND is defined even if cmake
2.8.12 is used.
also, the message() commands for debugging purpose are removed.
Signed-off-by: Kefu Chai <kchai@redhat.com>
we have switched from tmap to omap long ago.
but keep the server side implementation around, in case ancient
client is still using these tmap APIs.
also, tmap_update() is kept, because librbd is using it for v1 image
backward compatibility.
Signed-off-by: Kefu Chai <kchai@redhat.com>
It is particularly useful when running multiple rbd-mirror instances
in Active-Passive or Active-Active mode.
Signed-off-by: Mykola Golub <mgolub@suse.com>
include a patch so rocksdb can use libradospp instead of librados. will
upstream the patch and make it work for both pre-nautilus librados and
nautilus libradospp
Signed-off-by: Kefu Chai <kchai@redhat.com>
the goal is to decouple C++ API from C API, and to version them
differently, as they are targeting different consumers.
this allows us to change the C++ API and bumping up its soversion
without requiring consumer to recompile the librados client for
using the new librados. in this way, C++ API can move faster than
C API. for example, if bufferlist interface is changed for better
performance, and this breaks existing API/ABI, we can bump up
the C++ library's soversion, and and the C library's version unchanged
but ship the new librados's C binding. so the librados client linked
against librados's C library will be able to take advantage of
the improvement in C++ library. while the librados client
linked against C++ library won't break at runtime due to unresolved
symbol or changed structure layout.
this is massive change, the genereal idea is to
* split librados.cc into two source files: librados_c.cc and
librados_cxx.cc, the former for implementing C APIs, the later
for C++ APIs.
* extract the C++ API in librados into librados-cxx, the library
name will be libradospp. but we can change it before nautilus
is released.
* link these librados libraries with static libraries which it
depends on, so "-Wl,--exclude-libs,ALL" link flags can help
hide the non-public symbols.
* extract the tests exercising librados' C++ API into a different
source file named *_cxx.cc. for instance, to move the C++ tests
in aio.cc into aio_cxx.cc
* extract the shared helper functions which do not use any librados
or librados-cxx APIs into test_shared{.cc,h}. the "shared" here
means, *shared* by C++ and C tests.
* extract the test fixtures, i.e., the subclasses of testing::Test,
for testing C++ APIs into testcase_cxx.cc.
* update qa/workunits/rados/test.sh accordingly to add the splitted
tests
* update the consumers of librados to link against librados-cxx
instead, if they are using the C++ API.
Signed-off-by: Kefu Chai <kchai@redhat.com>
Henceforth, we'll require explicit `allow` caps for commands, or for the
config-key service. Blanket caps are no longer allowed for the
config-key service, except for 'allow *'.
Signed-off-by: Joao Eduardo Luis <joao@suse.de>
This is shown to corrupt otherwise healthy rocksdb databases. Rename to
make it clear that it is generally not safe to run and shoud only be used
as a last resort.
Signed-off-by: Sage Weil <sage@redhat.com>
* refs/pull/23069/head:
tests/libcephfs: add simple reclaim test
mds: check auth name before reclaiming session
mds: reclaim session before allowing mds to become active
mds: allow client to specify its session timeout
mds: initial code for client states reclaim
Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
expect_fail incorrectly unset '-e' option and if a consequent test
failed it did not abort the execution. And two typos in the namespace
tests were not detected due to this.
Signed-off-by: Mykola Golub <mgolub@suse.com>
in b38b8e980c, we changed the upper
limit of size of `config key` 's value to 64k, so we need to update
the test accordingly.
Fixes: http://tracker.ceph.com/issues/36260
Signed-off-by: Kefu Chai <kchai@redhat.com>
Sporadically the rbd-mirror fsx stress test would fail due to very
slow sync times due to overloaded clusters. Attempt to wait for all
images to be replicated before proceeding with the comparison.
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
* refs/pull/23187/head:
test: make rank argument mandatory when running journal_tool
cephfs-journal-tool: make "--rank" argument mandatory
cephfs-journal-tool: pass local arg vector for Journal actions
cephfs-journal-tool: dump to per rank output file wherever necessary
Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Two instances of fsstress clobber each other. Just build it in the local sandbox.
Fixes: http://tracker.ceph.com/issues/24177
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
Also, fix a bunch of quirky journal_tool invocations that pass
"--rank" argument as the command argument rather than passing it
as function argument.
Fixes: https://tracker.ceph.com/issues/24780
Signed-off-by: Venky Shankar <vshankar@redhat.com>
* refs/pull/24143/head:
qa/workunits/cephtool/test_kvstore_tool.sh: run test in ., not /tmp
Reviewed-by: Kefu Chai <kchai@redhat.com>
Reviewed-by: Radoslaw Zarzynski <rzarzyns@redhat.com>
5s -> 5m to gives us more leeway for when the mons are thrashing.
Also, *only* set this timeout when we expect a timeout. If we don't,
wait forever.
Signed-off-by: Sage Weil <sage@redhat.com>
* check if geom_gate can be loaded before doing the actual tests
Otherwise continuing does not make sense.
Major reason for this problem is due to mismatch between
kernel and module versions.
* After FreeBSD kernevel 1200078 ggate resizing is possible
So set the flag that resizing can be tested
* Only sudo commands that really need sudo
rbd-ggate list is available in regular user mode
* be a bit more verbose during testing and list the test purpose
* list-mapped is an option in rbd-nbd, not (yet) in rbd-ggate
Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>
* refs/pull/23240/head:
qa/suites/rados, qa/workunits/rados: Add suite/workunit for ceph-crash
add ceph-crash service
common/options: enable mgr 'crash' module by default
global/signal_handler: add 'done' file to signal crashdump is ready
Reviewed-by: Sage Weil <sage@redhat.com>
Selecting force peering on a single PG. In reality this probably induces
*2* interval changes.
Note that in the case of a single OSD cluster we can't actually force a
repeer on a single PG because the pg_temp code is pretty robust about
filtering out redundant or meaningless changes, so we can't pg_temp our
way into a new interval if there are no other OSDs to switch to and the
code also prevents an empty pg_temp.
Signed-off-by: Sage Weil <sage@redhat.com>
* refs/pull/22725/head:
qa/workunits/suites/blogbench.sh: use correct dir name
Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Josh Durgin <jdurgin@redhat.com>
Avoid sporadic failures in combination with msgr-failures/many.yaml,
where assert_locked() might take over 10 seconds.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
What we actually want is a purge, not a destroy. Destroy leaves the OSD
ID in used and allows it to be recreated. What ceph-volume wants is to
purge all trace of the failed OSD setup.
Signed-off-by: Sage Weil <sage@redhat.com>
