If the alignment is on a page boundary, or the allocation is big,
a separate buffer::raw goes faster. The rest of the time,
a raw_combined does.
Signed-off-by: Sage Weil <sage@redhat.com>
Removed the place holders in the open/close async state machines with
the newly added aio watch/unwatch methods.
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
Exclusive lock, object map, fast-diff, and deep-flatten have been
enabled by default for all new images.
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
"ceph --watch-debug" and "ceph tell mon.foo version" could connect
to different monitors, and there is chance that "ceph --watch-debug"
is not connected yet when "ceph tell" completes, and hence the former
fails to collect the cluster log including the "ceph tell" related
message. this renders test_mon_tell() unreliable. so, in
ceph_watch_start(), we should wait until the "ceph" cli connects to the
monitor and receives messages from it.
Fixes: #14910
Signed-off-by: Kefu Chai <kchai@redhat.com>
generate_object_name() uses std::ostringstream to build object name
and calls generate_object_prefix() (which uses another std::ostringstream),
making it quite inefficient and actually being one of limiting factors
in small-object benchmarks. New version that does everything on one char
array and within one function is up to 3,8x faster (in my testing bumping
throughput from 35MB/s to 40MB/s on rados bench rand -t 128 on 1KB objects).
Signed-off-by: Piotr Dałek <piotr.dalek@ts.fujitsu.com>
If a client recorded a maintenance op to the journal but crashed
before writing the op finish event, the image will be in an
inconsistent state. Therefore, once the end of the journal is
reached, attempt to apply all queued ops.
Fixes: #14822
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
If the start event is flagged as committed before the op is
actually executed, librbd won't be able to replay the event
should a crash occur.
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
Ignore errors that are to be expected when replaying
a maintenance op two or more times.
Fixes: #14822
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
Adding info about `rgw keystone verify ssl` to configuration reference,
also adding a note in rgw keystone to explain the usage.
Signed-off-by: Abhishek Lekshmanan <abhishek@suse.com>
msg: async: start over after failing to bind a port in specified range
Reviewed-by: Haomai Wang <haomai@xsky.com>
Reviewed-by: Kefu Chai <kefu@redhat.com>
The original report associated with this fix can be found at
http://tracker.ceph.com/issues/13002. I reopen one in case it
needs to be backported or something alike.
Fixes: #14928
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
This test is intended to check machine-readable output and usage, not
human-readable output.
Fixes: #14931
Signed-off-by: Josh Durgin <jdurgin@redhat.com>
pybind/rados: fix object lifetime issues and other bugs in aio
Reviewed-by: Mehdi Abaakouk <sileht@redhat.com>
Reviewed-by: Josh Durgin <jdurgin@redhat.com>
If bucket has requester-payer property set, aggregate the usage data
on it by the user that did the operation.
Signed-off-by: Yehuda Sadeh <yehuda@redhat.com>
Signed-off-by: Javier M. Mellid <jmunhoz@igalia.com>
This is awkward, if there are later users other than this test
and JournalingObjectStore, it'll be worth making it a bit less
clunky and error prone.
Signed-off-by: Samuel Just <sjust@redhat.com>
The rbd cli will warn about the deprecation when attempting to create
image format 1 images. librbd will log an error message when opening
a format 1 RBD image.
Signed-off-by: Jason Dillaman <dillaman@redhat.com>