* refs/pull/27619/head:
mgr/BaseMgrModule: run MonCommandCompletion on the finisher
mgr/BaseMgrModule: fix leak
Reviewed-by: Mykola Golub <mgolub@suse.com>
Reviewed-by: Tim Serong <tserong@suse.com>
currently, we use a `with_gate()` in `ProtocolV2::read_message()` for
ensuring that `this` (or `Connection` holding this protocol instance)
will outlive the continuation of `dispatcher.ms_dispatch()` which
references `this->dispatcher`. but we also pass a strong reference of
connection to dispatcher. in short, we have *two* safeguards for the
same purpose.
in this change, one of these safeguards is removed -- to pass the raw
pointer of `Connection` to `Dispatch::ms_dispatch()`. the reason why
the `with_gate()` is kept is that, if we have removed `with_gate()` in
Protocol, we need to
1. let `Dispatcher::ms_dispatch()` return `void`, as it should not block
any succeeding calls.
2. add a `with_gate()` in `Dispatcher::ms_dispatch()` to ensure that
`this` is alive during the lifecycle of the continuation(s) in
`Dispatcher::ms_dispatch()`.
Signed-off-by: Yingxin Cheng <yingxincheng@gmail.com>
Signed-off-by: Kefu Chai <kchai@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>
Also tweaked the existing writeback cache test permutations to
specify the use of the writeback cache policy.
Fixes: http://tracker.ceph.com/issues/39386
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
The completion takes the ActivePyModules::lock, but the caller may be
holding an internal lock for Objecter and/or MonClient. Avoid deadlock
by putting the completion on our finisher.
Signed-off-by: Sage Weil <sage@redhat.com>
When exec: rbd-nbd map rbd/image --device /dev/image
The error message is:
rbd-nbd: failed to open device: /dev/image.
In fact, it should print:
rbd-nbd: invalid device path: /dev/image (expected /dev/nbd{num})
Meantime change help which make more understand.
Fixes: http://tracker.ceph.com/issues/39269
Signed-off-by: Jianpeng Ma <jianpeng.ma@intel.com>
MDSMonitor assigns a booting MDS immediately to standby. From there, the
standby may be promoted to standby-replay or some state leading to active.
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
* refs/pull/27375/head:
rgw: pass mostly_omap flag when opening/creating pools
rgw/rgw_rados: pass mostly_omap flag when opening/creating pools
rgw: move rgw_init_ioctx() to rgw_tools.cc
Reviewed-by: Casey Bodley <cbodley@redhat.com>
The destructor for ImageCtx attempts to access librados::IoCtx objects
that might have been destroyed immediately after the async open (failure)
and close actions completed.
Fixes: http://tracker.ceph.com/issues/39031
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
stop command can be used to force stopping a specified osd daemon, e.g.,
you don't have to pre-figure out where it located.
Signed-off-by: xie xingguo <xie.xingguo@zte.com.cn>
This is an opt-in wrapper around Message inheritance to prevent
undesired get/put calls on the object. Instead, you must/should use a
smart ptr to adjust the reference count.
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
Signed-off-by: Kefu Chai <kchai@redhat.com>
