J. Eric Ivancich [Wed, 28 Jul 2021 17:52:29 +0000 (13:52 -0400)]
rgw: user stats showing 0 value for "size_utilized" and "size_kb_utilized" fields
When accumulating user stats, the "utilized" fields are not looked
at. Updates RGWStorageStats::dump so it only outputs the "utilized"
data if they're updated.
Kajetan Janiak [Wed, 18 Nov 2020 10:42:07 +0000 (11:42 +0100)]
rgw: disable prefetch in rgw_file
Each call to rgw_read (rgw_file.cc) invokes three calls to RGWRados::get_obj_state with s->prefetch_data=true. It results in great read amplification. If length argument in rgw_read call is smaller than rgw_max_chunk_size, then the amplification is threefold.
Adam Kupczyk [Sat, 13 Nov 2021 10:28:18 +0000 (11:28 +0100)]
os/bluestore: Fix omap upgrade to per-pg scheme
This is fix to regression introduced by fix to omap upgrade: https://github.com/ceph/ceph/pull/43687
The problem was that we always skipped first omap entry.
This worked fine with objects having omap header key.
For objects without header key we skipped first actual omap key.
Fixes: https://tracker.ceph.com/issues/53260 Signed-off-by: Adam Kupczyk <akupczyk@redhat.com>
(cherry picked from commit 65a3f374aa1c57c5bb9401e57dab98a643b4360a)
Those functions should return `None` if 0 or more than 1 item is returned.
The current name of these functions are confusing and can lead to thinking that
we just want the first item returned, even though it returns more than 1
item, let's rename them to `get_single_pv()`, `get_single_vg()` and
`get_single_lv()`
Mark Kogan [Mon, 15 Nov 2021 15:50:49 +0000 (15:50 +0000)]
rgwi/beast: stream timer with duration 0 disables timeout
fixes all S3 operations failing with:
`2021-11-15T15:46:05.992+0000 7ffee17fa700 20 failed to read header: Bad file descriptor`
when `--rgw_frontends="beast port=8000 request_timeout_ms=0"`
Casey Bodley [Thu, 11 Nov 2021 17:01:06 +0000 (12:01 -0500)]
rgw/beast: reference count Connections for timeout_handler
resolves a use-after-free in the timeout_handler, where a timeout fires
and schedules the timeout_handler for execution, but the coroutine exits
and destroys the socket before asio executes the timeout_handler
timeout_handler now holds a reference on the Connection to extend its
lifetime
now that the Connection is allocated on the heap, we can include the
parse_buffer in this memory instead of allocating it separately
Casey Bodley [Sat, 30 Oct 2021 23:47:02 +0000 (19:47 -0400)]
rgw/beast: replace beast::tcp_stream with manual timeouts
remove the beast::tcp_stream wrapper from the socket, and track timeouts
manually with a timeout_timer. this timer uses ceph's coarse_mono_clock
which is cheaper to sample than std::chrono::steady_clock
Casey Bodley [Mon, 1 Nov 2021 17:14:16 +0000 (13:14 -0400)]
spawn: use explicit strand executor
the default spawn::yield_context uses the polymorphic boost::asio::executor
to support any executor type
rgw's beast frontend always uses the same executor type for these
coroutines, so we can use that type directly to avoid the overhead of
type erasure and virtual function calls
Cherry-pick notes:
- src/rgw/rgw_d3n_cacherequest.h doesn't exist in Octopus
- src/rgw/rgw_sync_checkpoint.cc doesn't exist in Octopus
- conflicts due to rename of structs after Octopus
- conflicts due to macro for conditional inclusion of beast context in Octopus
The current check only allows to request an OSD id that exists but
marked as 'destroyed'.
With this small fix, we can now use `--osd-id` with an id that doesn't
exist at all.
The marker was not working correctly as segments of the bucket index
were listed to shut down any incomplete multipart uploads. This fixes
the marker, so it's maintained properly across iterations.
Igor Fedotov [Thu, 15 Jul 2021 12:10:14 +0000 (15:10 +0300)]
os/bluestore: fix improper offset calculation when repairing.
While repairing misreferenced blobs BlueStore could improperly calculate
an offset within a blob being fixed. This could happen when single
physical extent has been replaced by multiple ones - the following
pextent (if any in the current blob) would be treated with the improper offset within the blob. Offset calculation didn't account for each of that new pextents but the last one only.
Fixes: https://tracker.ceph.com/issues/51682 Signed-off-by: Igor Fedotov <ifedotov@suse.com>
(cherry picked from commit ca4b6675fc3fd2f4cadad58044c97c5bb23d5938)
Igor Fedotov [Thu, 15 Jul 2021 11:16:39 +0000 (14:16 +0300)]
test/objectstore/bluestore_types: add map_bl test case
Along with the basic bluestore_blob_t::map_any functionality
verification this UT shows how invalid offset might appear in
https://tracker.ceph.com/issues/51682
so AvlAllocator can switch from the first-first mode to best-fit mode
without walking through the whole space map tree. in the
highly-fragmented system, iterating the whole tree could hurt the
performance of fast storage system a lot.
the idea comes from openzfs's metaslab allocator.
Signed-off-by: Kefu Chai <kchai@redhat.com>
(cherry picked from commit 5a26875049d13130ffe5954428da0e1b9750359f) Signed-off-by: Mauricio Faria de Oliveira <mfo@canonical.com>
Conflicts:
src/common/options/global.yaml.in:
- Moved new option into src/common/options.cc
so AvlAllocator can switch from the first-first mode to best-fit mode
without walking through the whole space map tree. in the
highly-fragmented system, iterating the whole tree could hurt the
performance of fast storage system a lot.
the idea comes from openzfs's metaslab allocator.
Signed-off-by: Kefu Chai <kchai@redhat.com>
(cherry picked from commit 40f05b971f5a8064cf9819f80fc3bbf21d5206da) Signed-off-by: Mauricio Faria de Oliveira <mfo@canonical.com>
Conflicts:
src/common/options/global.yaml.in
- Moved new option into src/common/options.cc
Kefu Chai [Wed, 2 Jun 2021 08:31:18 +0000 (16:31 +0800)]
os/bluestore/AvlAllocator: use cbit for counting the order of alignment
no need to calculate the alignment first, cbits() would suffice. as it
counts the first set bit and the follow 0's in a number. the result
is identical to the cbit(alignment of that number).
Signed-off-by: Kefu Chai <kchai@redhat.com>
(cherry picked from commit 573cbb796e8ba2f433caa308925735101a8161a6) Signed-off-by: Mauricio Faria de Oliveira <mfo@canonical.com>
before this change AvlAllocator::_block_picker() is used by both the
best-fit mode and first-fit mode. but since we cannot achieve the
locality by searching in the area pointed by curosr in best-fit mode,
we just pass a dummy cursor to AvlAllocator::_block_picker() when
searching in the best-fit mode.
but since the range_size_tree is already sorted by the size of ranges,
if _block_picker() fails to find one by the size, we should just give
up right away, and instead try again using a smaller size.
after this change, instead of sharing AvlAllocator::_block_picker()
across both the first-fit mode and the best-fit mode, this method
is specialize to two different variants: one for first-fit, and the
simplified one for best-fit.
Signed-off-by: Kefu Chai <kchai@redhat.com>
(cherry picked from commit 4837166f9e7a659742d4184f021ad12260247888) Signed-off-by: Mauricio Faria de Oliveira <mfo@canonical.com>
Adam Kupczyk [Wed, 19 May 2021 10:49:37 +0000 (12:49 +0200)]
os/bluestore: Improve _block_picker function
Make _block_picker function scan (*cursor, end) + (begin, *cursor) instead of (*cursor, end) + (begin, end).
The second run over range (*cursor, end) could never yield any results.
Signed-off-by: Adam Kupczyk <akupczyk@redhat.com>
(cherry picked from commit c732060d3e3ef96c6da06c9dde3ed8c064a50965) Signed-off-by: Mauricio Faria de Oliveira <mfo@canonical.com>
Igor Fedotov [Tue, 9 Feb 2021 15:29:01 +0000 (18:29 +0300)]
os/bluestore: cap omap naming scheme upgrade transactoin.
We shouldn't use single per-onode transaction for such an upgrade when onode's omap list is huge. This results in similarly sized WAL/SST files which are inefficient, might cause high memory usage and sometimes error-prone.
Fixes: https://tracker.ceph.com/issues/49170 Signed-off-by: Igor Fedotov <ifedotov@suse.com>
(cherry picked from commit e897fa243c1dd38329733b452872616023f14ac8)
Conflicts (caused by lack of per-pg omap scheme):
src/os/bluestore/BlueStore.cc
src/os/bluestore/BlueStore.h
src/os/bluestore/bluestore_types.h
Casey Bodley [Thu, 4 Nov 2021 16:12:04 +0000 (12:12 -0400)]
qa/rgw: octopus branch targets ceph-octopus branch of java_s3tests
this commit is applied directly to the octopus branch instead of a
backport from master, because it targets the ceph-octopus branch instead
of the ceph-master branch on master
test/rgw: fix use of poll() with timers in unittest_rgw_dmclock_scheduler
the AsyncScheduler uses an asio timer to dispatch work to its executor
with an optional delay. when no delay is requested, it waits on the
timer with an expiration time in the past (crimson::dmclock::TimeZero)
tests are failing here because poll() is returning without executing the
handlers of those expired timers
asio implements these timers with timerfd and epoll. debugging with
strace, i see that these timers armed with timerfd_settime() are not
always immediately ready according to epoll_wait():
Kefu Chai [Sat, 30 Oct 2021 03:18:17 +0000 (11:18 +0800)]
admin/doc-requirements.txt: pin Sphinx at 3.5.4
* pin Sphinx at 3.5.4
* pin docutils at 0.18
at least the combination of these two versions
is known to compile.
to address the bug reported at
https://sourceforge.net/p/docutils/bugs/431/
the backtrace looks like:
/home/jenkins-build/build/workspace/ceph-pr-docs/build-doc/virtualenv/lib/python3.8/site-packages/sphinx/util/docutils.py:285:
RemovedInSphinx30Warning: function based directive support is now
deprecated. Use class based directive instead.
warnings.warn('function based directive support is now deprecated. '
Exception occurred:
File
"/home/jenkins-build/build/workspace/ceph-pr-docs/build-doc/virtualenv/lib/python3.8/site-packages/docutils/writers/html5_polyglot/__init__.py",
line 445, in section_title_tags
if (ids and self.settings.section_self_link
AttributeError: 'Values' object has no attribute 'section_self_link'
Kefu Chai [Tue, 17 Aug 2021 07:53:51 +0000 (15:53 +0800)]
mgr/dashboard/api: set a UTF-8 locale when running pip
ansible-core started to include files whose filenames are encoded in
non-ascii characters, so we have to use a more capable encoding for the
locale in order to install this package. otherwise we'd have following
error:
Collecting ansible-core<2.12,>=2.11.3
Using cached ansible-core-2.11.4.tar.gz (6.8 MB)
ERROR: Exception:
Traceback (most recent call last):
File "/tmp/tmp.fX76ASIrch/venv/lib/python3.8/site-packages/pip/_internal/cli/base_command.py", line 173, in _main
status = self.run(options, args)
...
File "/tmp/tmp.fX76ASIrch/venv/lib/python3.8/site-packages/pip/_internal/utils/unpacking.py", line 226, in untar_file
with open(path, "wb") as destfp:
UnicodeEncodeError: 'latin-1' codec can't encode characters in position 137-140: ordinal not in range(256)
The ragodgw-admin bucket limit check command has a bug in
octopus. Since we do not clear the bucket list before
list_buckets() returns the next max_entries, they are appended
to the existing list and we end up counting the first ones again.
This bug is triggered if bucket count exceeds max_entries and
causes duplicates in the output of radosgw-admin bucket limit check.
The fix clears the buckets structure before the list_buckets()
populates it again with the next lot of buckets to iterate through.
Mykola Golub [Fri, 20 Mar 2020 16:19:25 +0000 (16:19 +0000)]
ceph-erasure-code-tool: new tool to encode/decode files
E.g. it may be useful as a last resort when recovering an object from
a damaged PG: extract the encoded object chunks from the PG shards
with ceph-objectstore-tool and then decode with ceph-erasure-code-tool.
It also has functionality similar to what ceph_erasure_code test provides.
Kefu Chai [Thu, 10 Jun 2021 12:19:09 +0000 (20:19 +0800)]
tasks/ceph_manager: ignore EACCES when waiting for quorum
mon_tick_interval is 5 seconds by default. monitors update their
rotating keys every mon_tick_interval. before monitors forms a
quorum, the auth requests from clients are put into the wait list.
these requests are re-enqueued once the monitors form a quorum. but
there is a small window of mon_tick_interval, before they are able
to serve the auth requests even after their claim to be able to
server requests. if these re-enqueued requests happen to be served
in this window, and if authx is enabled, they will be greeted with
errors like
handle_auth_bad_method server allowed_methods [2] but i only support [2]
in the case of ceph cli, the error would look like:
[errno 13] RADOS permission denied (error connecting to the cluster)
so, to address this issue, the EACCES error is ignored when waiting
for a quorum.
ceph-monstore-tool: use a large enough paxos/{first,last}_committed
so the rebuild paxos transaction won't be overwritten by the ones
created before recovery completes.
when the quorum is recovering, the leader will collect the paxos
transactions from peons. if the quorum accept the proposal for setting
the fingerprint, the peon will update the monitor with the paxos
transaction with a newer "last_committed" than the one created using
update_paxos() in ceph_monstore_tool.cc. the latter "last_committed" is
always 0.
so, to avoid this extra paxos proposal obsoleting the "rebuilding" paxos
transaction, we use a large enough number for {first,last}_committed.
projects / ceph.git / log