- those old pg settings are no longer lost as they go in the OSDMap now
- add a note about pools that were mid-create
Signed-off-by: Sage Weil <sage@redhat.com>
In particular, we need to rebuild the created_pools set or else the mon
will try to recreate every PG in the system. On an aged cluster, that
is extremely expensive, and will break things.
Fixes: http://tracker.ceph.com/issues/36306
Signed-off-by: Sage Weil <sage@redhat.com>
cmake: do not use plain target_link_libraries(rgw_a ...)
Reviewed-by: Laura Paduano <lpaduano@suse.com>
Reviewed-by: Kai Wagner <kwagner@suse.com>
Reviewed-by: Casey Bodley <cbodley@redhat.com>
this addresses following error:
CMake Error at src/rgw/CMakeLists.txt:178 (target_link_libraries):
The plain signature for target_link_libraries has already been used
with
the target "rgw_a". All uses of target_link_libraries with a target
must
be either all-keyword or all-plain.
The uses of the plain signature are here:
* src/rgw/CMakeLists.txt:168 (target_link_libraries)
Signed-off-by: Kefu Chai <kchai@redhat.com>
mgr broadcast "notify()" to call registered plugins upon finishing a
send_command() request. and restful plugin prints a warning message if
the request is not issued by itself. apparently, there is good chance
that the finished request is sent by other mgr plugins, in that case,
restful plugin always prints the warning messages like:
"2018-10-09 14:52:07.818 7fbc4d8c5700 1 mgr[restful] Unknown request
''
this is misleading and annoying. so, in this change
* add a prefix of "restful.module" before that tag used for identifying
a request. this prefix helps to differentiate the requests sent by
restful plugin from other requests.
* do not print warning message if none running request matches the given tag
* break at seeing the first request matches the given tag, because
the tags used by restful plugin are unique.
* do not str(self.id) before composing the tag, as it is a `str`
already, see CommandsRequest.__init__()
* early return to reduce the indent level
Fixes: http://tracker.ceph.com/issues/36374
Signed-off-by: Kefu Chai <kchai@redhat.com>
Make sure the mbuf and ethdev libraries are created as they export
required symbols.
Fixes: http://tracker.ceph.com/issues/36341
Signed-off-by: Brad Hubbard <bhubbard@redhat.com>
It is easy for common users to understand why their clients not working
due to blacklisted state.
Fixes: http://tracker.ceph.com/issues/36352
Signed-off-by: Zhi Zhang <zhangz.david@outlook.com>
This caused DPP to be threaded through the
verify_{objecy/bucket/user}_permissions functions
in rgw_common.cc and change all of the calls to
them in rgw_op.cc and in other places.
The logging in STS engine authorize was
changed too.
Signed-off-by: Ali Maredia <amaredia@redhat.com>
Apparently 15m is not long enough for some workunits like fsstress.
Fixes: http://tracker.ceph.com/issues/36365
Signed-off-by: Patrick Donnelly <pdonnell@redhat.com>
pass DPP into authenticate functions for auth
engines and change any of the logging inside of
them.
Next up is passing DPP and changing the douts
in the helper functions called in authenticate.
Signed-off-by: Ali Maredia <amaredia@redhat.com>
Pass a DPP into RGWOp::verify_request that leads
through RGWHandler -> RGWHandler_REST->
RGW::Auth_S3::authorize().
Change authenticate() in all the Handler classes
to take a DPP
Signed-off-by: Ali Maredia <amaredia@redhat.com>
The libcls_*.so libraries are runtime dependencies of the OSD (and RGW).
Building the vstart target doesn't build these libraries, rendering some
functionalities unusable. This goes unnoticed when building the entire
source tree.
Signed-off-by: Mohamad Gebai <mgebai@suse.com>
