If there is a bug preventing rm from completing, the workunit will get stuck.
Fixes: http://tracker.ceph.com/issues/36184
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
Otherwise QA sits forever waiting for the kclient to umount when there is a
problem.
Fixes: http://tracker.ceph.com/issues/36184
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
Otherwise the command will hang if the mount is broken.
Fixes: http://tracker.ceph.com/issues/36184
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
in b38b8e980c, we changed the upper
limit of size of `config key` 's value to 64k, so we need to update
the test accordingly.
Fixes: http://tracker.ceph.com/issues/36260
Signed-off-by: Kefu Chai <kchai@redhat.com>
With this command, the MDS would request clients to release
caps followed by trimming its own cache and a journal flush.
The command accepts a timeout to wait for clients to respond
to session recall and flush messages.
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
Signed-off-by: Rishabh Dave <ridave@redhat.com>
Signed-off-by: Venky Shankar <vshankar@redhat.com>
Uses the new ``CephDiskDevice`` to look into PARTLABEL from both lsblk
(the default) falling back to blkid, which in some cases has the right
value when lsblk fails.
Signed-off-by: Alfredo Deza <adeza@redhat.com>
When using direct function calls to implement the V1 state machine,
we might fill up the stack due to state machine cycles.
With continuations we fix this problem by limiting the depth of
the call stack.
Fixes: http://tracker.ceph.com/issues/36167
Signed-off-by: Ricardo Dias <rdias@suse.com>
Pgs show the unknown state, when the mgr has just started, use the command ceph pg dump pgs_brief\pgs\...,
because the pgs state has not been reported from the Osd to the Mgr service,
we need to add this state for pgs show. in this way,our management system can judge whether to filter
the unknown pg state based on this flag.
Fixes: http://tracker.ceph.com/issues/25103
Signed-off-by: huanwen ren <ren.huanwen@zte.com.cn>
in this change, along with LockPolicy, src/common/lock_* are completely
removed.
instead of using LockPolicy based template specialization of LockMutex,
etc, it would be simpler if we can just rely on WITH_SEASTAR
preprocessor macro to tell if we are compiling code for crimson or not.
but please bear in mind, we cannot link against the plain libceph-common
in crimson anymore.
Signed-off-by: Kefu Chai <kchai@redhat.com>
for the sake of completeness, also pave the road to ditch
SharedMutex<LockPolicy>. please note, in code used by crimson, we are
not using shared_mutex or SharedMutex yet. so we are not adding
the dummy shared_mutex for WITH_SEASTAR.
Signed-off-by: Kefu Chai <kchai@redhat.com>
When connection stuck into odd state, we need to let connection timeout.
If send_message could update last_active, it won't play the role as
expected.
So we move last_active updated to the place actually should.
Signed-off-by: Haomai Wang <haomai@xsky.com>
Signed-off-by: Kefu Chai <kchai@redhat.com>
Introducing Clay codes to ceph. These codes are optimal
in terms of network, disk traffic utilized during repair of a lost chunk.
This feature also provides gradual increase in network, disk traffic with
respect to the number of lost chunks.
Authors: Myna, Elita.
Fixes: http://tracker.ceph.com/issues/19278
Signed-off-by: Myna V <mynaramana@gmail.com>
