Sometimes ERROR log entries do not have prior_version. Also,
they aren't actually updates and don't affect the object.
Fixes: http://tracker.ceph.com/issues/20274
Signed-off-by: Sage Weil <sage@redhat.com>
On NVMe we can write really fast... fast enough that we may fill up before
we realize we are nearly full. It's a problem here only because our
devices are so small that the OSDs fill up proportionally quickly.
Signed-off-by: Sage Weil <sage@redhat.com>
Centos btrfs hits ENOSPC all the time :(, and now xenial is doing
it too.
See http://tracker.ceph.com/issues/20169
Signed-off-by: Sage Weil <sage@redhat.com>
osd, messages/MOSDPing: bunch of fixes related to ping inflation
Reviewed-by: Sage Weil <sage@redhat.com>
Reviewed-by: Greg Farnum <gfarnum@redhat.com>
cmake: do not add dependencies to INTERFACE library on cmake < 3.3
Reviewed-by: Casey Bodley <cbodley@redhat.com>
Reviewed-by: Joao Eduardo Luis <joao@suse.de>
This removes the default filename, by the way. We also work around a
problem with make_server where it sets up the socket to listen before
checking for the cert, thereby making it problematic to rebind to the
port shortly thereafter when we do have a socket. (SO_REUSEADDR would
be appropriate but there doesn't seem to be an easy way to make
make_server use it.)
Signed-off-by: Sage Weil <sage@redhat.com>
This implements a create-self-signed-cert command for the ceph-mgr
restful interface.
It is designed so that it will try to restart the module once the cert
is created.
Signed-off-by: Boris Ranto <branto@redhat.com>
well since flat_map is a vector under the hood it should be faster for
such small sizes, also renaming tags to a more understandable tag_map.
Signed-off-by: Abhishek Lekshmanan <abhishek@suse.com>
A very basic support for the s3:{get/put/delete}objecttagging and
related versions for object versioning
Signed-off-by: Abhishek Lekshmanan <abhishek@suse.com>
- New api for get/put/delete object tags
- PUT Object supports x-amz-tagging
- POST API supports x-amz-tagging
This is similar to AWS S3's get/put object tags, tags are stored as an
xattr under the user.rgw.tags key, and are restricted to max 10 tags per
object and key size of 128 and a val size of 256. The API does not reuse
the existing `x-amz-meta` tags.
For PUT obj to optionally accept `x-amz-tagging` and this will end up
storing the tags under x-amz-tagging xattr, the input is expected to be
in the url encoded kv format, we return InvalidTagError if we cannot
parse the tags
A future TODO: use the same xml parser for put and post apis
Signed-off-by: Abhishek Lekshmanan <abhishek@suse.com>
Conflicts:
src/rgw/rgw_rest_s3.cc
src/rgw/rgw_tag.cc
mostly trivial conflicts from AWSv4 rework, importantly url_decode's
signature was changed for the nicer, and moved rgw_tag.cc to the
newer url_decode format
Conflicts:
src/rgw/rgw_common.h
conflict as another PR that used up the error code before this was
merged
Currently users can't use ceph admin command to get db stats when running
bluestore fio plugin. This patch is to print db stats when rocksdb_perf
is set in config file.
Signed-off-by: Xiaoyan Li <xiaoyan.li@intel.com>
Otherwise, path string like #10000000000/xxx may get saved into
inode_t::stray_prior_path.
Fixes: http://tracker.ceph.com/issues/20340
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Create an mClock priority queue, which can in turn be used for two new
implementations of the PG shards operator queue. The first
(mClockOpClassQueue) prioritizes operations based on which class they
belong to (recovery, scrub, snaptrim, client op, osd subop). The
second (mClockClientQueue) also incorporates the client identifier, in
order to promote fairness between clients.
In addition, also remove OpQueue's remove_by_filter and all possible
associated subclass implementations and tests.
Signed-off-by: J. Eric Ivancich <ivancich@redhat.com>
Sending any SLO/DLO range requests will retrun InvalidRange error. SLO/DLO should call range_to_ofs after got object size instead of read_op.prepare() to deal with negative offsets.
Signed-off-by: Shasha Lu <lu.shasha@eisoo.com>
