if map client_caps has not key client, client_caps[client] will insert key client with null value into the map.
Fixes: https://tracker.ceph.com/issues/40939
Signed-off-by: XiaoGuoDong2019 <xiaogd@inspur.com>
* refs/pull/28847/head:
doc/mgr/telemetry: update
mgr/telemetry: move contact info to an 'ident' channel
mgr/telemetry: accept channel list to 'telemetry show'
mgr/telemetry: always generate new report for 'telemetry show'
mgr/telemetry: add 'device' channel and call out to devicehealth module
mgr/telemetry: add telemetry channel 'device'
mgr/telemetry: add separate channels
Reviewed-by: Dan Mick <dmick@redhat.com>
* refs/pull/29221/head:
mon/OSDMonitor: do not clobber pending xinfo during boot
osd/OSDMap: add is_dead() helper
mon/OSDMonitor: 'osd down ... --definitely-dead'
osd: send MOSDMarkMeDead when we are marked down
mon/OSDMonitor: implement MOSDMarkMeDead handling
mon/OSDMonitor: set dead_epoch for immediate failure (ECONNREFUSED)
osd/OSDMap: add dead_epoch to xinfo
Reviewed-by: Samuel Just <sjust@redhat.com>
* refs/pull/28874/head:
ceph-mds: dump all info of ceph_file_layout, InodeStoreBase, frag_info_t, inode_t
Reviewed-by: Zheng Yan <zyan@redhat.com>
Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
* refs/pull/29166/head:
mds: Reorganize class members in Capability header
Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Jos Collin <jcollin@redhat.com>
* refs/pull/28860/head:
mds: Move scalar type initializations of CDir data members to its header
mds: Reorganize class members in CDir header
Reviewed-by: Patrick Donnelly <pdonnell@redhat.com>
Reviewed-by: Jos Collin <jcollin@redhat.com>
If we have a pending xinfo change and also process a boot message, we
should not throw out the pending xinfo change.
This triggers when you mark an osd down and it sends both a MOSDMarkMeDead
and a new MOSDBoot message in quick succession: the first message sets
dead_epoch but the boot message processing clobbers it.
Signed-off-by: Sage Weil <sage@redhat.com>
Like 'osd down', but we also set dead_epoch. Should only be used if we
know the process/host is already completely dead and unresponsive.
Signed-off-by: Sage Weil <sage@redhat.com>
* refs/pull/29111/head:
qa/mgr/progress: Update the test suite for progress module
Reviewed-by: Josh Durgin <jdurgin@redhat.com>
Reviewed-by: Kefu Chai <kchai@redhat.com>
Allow updates to the dead_epoch in osd_xinfo, so that the OSD can inform
us that they know they are down (as of a particular epoch).
Signed-off-by: Sage Weil <sage@redhat.com>
The 'ceph' CLI will duplicate commands within teuthology to test
the MONs idempotency. This shouldn't be required for the MGR module,
but we can keep a fixed set of completed tests to handle this
possible command replay.
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
This callback can be used to track progress and also to attempt to cancel
the operation while it's in-progress by returning a negative error code
from the callback.
Signed-off-by: Jason Dillaman <dillaman@redhat.com>
This updates describe_service() to include nfs and iscsi services
(deepsea internally refers to these as "ganesha" and "igw" roles).
Additionally, if deepsea sets any of container_id, service, version,
rados_config_location, service_url, status or status_desc, these will
come through now too.
This relies on https://github.com/SUSE/DeepSea/pull/1606 for the new
functionality, but if run against an older version of deepsea, it will
continue to operate as it did before, it just won't include
rados_config_location or service_url data.
I've also updated the caching in describe_service() so that if the
cache needs filling, it just requests all services from DeepSea, caches
everything, then does filtering on the results. Previously, if the
cache was empty and only one service type was requested, the cache
would then *only* include that service and no other items, so subsequent
requests for different service types would return nothing until the
cache expired.
Fixes: https://tracker.ceph.com/issues/39095
Signed-off-by: Tim Serong <tserong@suse.com>
`ceph orchestrator device ls` already supports --refresh to force
a reload, we should support this for `ceph orchestrator service ls`
as well.
Signed-off-by: Tim Serong <tserong@suse.com>
use a smart pointer instead of using 'new' to resolve a memory leak .
Fixes: https://tracker.ceph.com/issues/40936
Signed-off-by: XiaoGuoDong2019 <xiaogd@inspur.com>
fix klocwork issues:
function 'strcpy' does not check buffer boundaries but outputs to buffer 'watcher.addr' of fixed size (256）
Signed-off-by: songweibin <song.weibin@zte.com.cn>
