The librbd API is capped at 24 characters for expressing the
object prefix for data blocks (including trailing null byte).
Fixes: http://tracker.ceph.com/issues/16887
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
Commit c38e3cbb6f introduced a second
compiled version of src/civetweb/src/civetweb.c, but did not pass the
configuration header (civetweb/include/civetweb_conf.h).
As a result, USE_IPV6 was not defined when it was compiled, and that
copy was included into the radosgw binary. This caused breakage for the
civetweb frontend when used with IPv6:
rgw frontends = civetweb port=[::]:7480
Reintroduce the header so that civetweb is compiled correctly again.
Fixes: http://tracker.ceph.com/issues/16928
Backport: jewel
Signed-off-by: Robin H. Johnson <robin.johnson@dreamhost.com>
Describe how to configure logging and the admin socket for debugging RBD
libvirt clients.
Thanks to Jason Dillman <jdillama@redhat.com> on ceph-devel for this
information.
Signed-off-by: Ken Dreyer <kdreyer@redhat.com>
From the Linux manual:
MAP_ANONYMOUS
The mapping is not backed by any file; its contents
are initialized to zero. The fd and offset arguments are
ignored; however, some implementations require fd to be -1
if MAP_ANONYMOUS (or MAP_ANON) is specified, and portable
applications should ensure this.
FreeBSD is such a case, it wil just return an error.
- Used in:
src/test/on_exit.cc
src/test/system/cross_process_sem.cc
Signed-off-by: Willem Jan Withagen <wjw@digiware.nl>
Hard code transaction size to be 300 is not good. Since we already
introduce osd_target_transaction_size for the same purpose, we can
use it directly here.
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
When _have_pg() returns true, we drop the pg_map_lock
and pg->lock() simultaneously. So theoretically the
subsequent call to _lookup_lock_pg() can return NULL,
although the chance is rare.
Also, since we are going to call _lookup_lock_pg() anyway,
we can call it directly, thus the _have_pg() method is
totally unnecessary.
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
Basically we don't allow a transaction to get too big, e.g., to
avoid contiguously occupation of CPU.
We break and queue the transaction here whenever we have collected
osd_target_transaction_size temp objects.
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
Everyone should upgrade through Firefly -> Hammer -> etc to pick up the newer
package names.
Signed-off-by: Ken Dreyer <kdreyer@redhat.com>
Signed-off-by: Nathan Cutler <ncutler@suse.com>
libcommon.a is linked statically to librados and libcephfs. When the global
destructors for these libraries are called malloc code can report "double free
or corrpution" as the same resource may be deleted twice (global static
non-POD variable types). This test attempts to detect this issue.
Signed-off-by: Brad Hubbard <bhubbard@redhat.com>
utime_t convert double in MDBalancer::tick(),
so g_conf->mds_bal_sample_interval should be set OPT_DOUBLE
Signed-off-by: huanwen ren <ren.huanwen@zte.com.cn>
The destructor of OSD will delete class_handler arbitrarily
and there are cases we'll fail to start an osd before we
load the class_handler correctly.
So it is safe to initialize class_handler to NULL as to
avoid access violation.
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
According to Linux Man pages:
ECANCELED Operation canceled (POSIX.1)
So ECANCELED is the preferred one.
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
The agent_timer.shutdown() method will call cancel_all_events()
within in a more safe and elegant way, so we don't have to
call it explicitly here.
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
src/test/ceph_objectstore_tool.py depends on an implicit "-k" when not
creating a new cluster. (Deleting the conf file without -n makes no
sense anyway.)
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
