Both stupid and bitmap allocator returs -ENOSPC if they're
unable to allocate any space. Existing callers aren't always
respect this - hence doing some cleanup.
Signed-off-by: Igor Fedotov <ifedotov@suse.com>
* remove duplicated information in message logs,
because connection prefix already contains information about:
"[my_name(my_logic_name) my_addr >> peer_name peer_addr]"
* unify message logs for both v1 and v2;
* continue to use familiar identifiers to highlight message logs:
- received messages identified by "<== #seq === msg (type)";
- sent messages identified by "--> #seq === msg (type)";
Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>
Defind log levels in V2 Protocol:
* error level, something error that cause connection to terminate:
- fatal errors;
- bugs;
* warn level: something unusual that identifies connection fault or replacement:
- unstable network;
- incompatible peer;
- auth failure;
- connection race;
- connection reset;
* info level, something very important to show connection lifecycle,
which doesn't happen very frequently;
* debug level, important logs for debugging, including:
- all the messages sent/received (-->/<==);
- all the frames exchanged (WRITE/GOT);
- important fields updated (UPDATE);
- connection state transitions (TRIGGER);
* trace level, trivial logs showing:
- the exact bytes being sent/received (SEND/RECV(bytes));
- detailed information of sub-frames;
- integrity checks;
- etc.
Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>
* use standard entity_type_t(uint8_t) in favor of peer_type_t(int);
* use standard ceph_entity_type_name() to print where possible;
* use peer_name to encapsulate peer_type and peer_id in Connection.h;
* clean related interfaces in Messenger and Connection;
* print peer_name in Connection logging prefix;
Signed-off-by: Yingxin Cheng <yingxin.cheng@intel.com>
This reverts commit d92117d8ea7018c533cc422e31c09d7e45a0793a
at the rason for it has been fixed.
Signed-off-by: Radoslaw Zarzynski <rzarzyns@redhat.com>
Use the 'cd-loading-panel' element to display the loading progress.
Fixes: https://tracker.ceph.com/issues/40427
Signed-off-by: Volker Theile <vtheile@suse.com>
mgr/dashboard: Form fields do not show error messages/hints.
Reviewed-by: Ricardo Marques <rimarques@suse.com>
Reviewed-by: Tiago Melo <tmelo@suse.com>
for MOSDOp, get_min_epoch() is an alias of get_map_epoch(), but
get_min_epoch() is symantically more correct.
Signed-off-by: Kefu Chai <kchai@redhat.com>
differentiate WRITE from WRITEALL.
and handles differnt cases of `os.oi.truncate_seq` caused by
`write_trunc` and/or `trunc` ops.
Signed-off-by: Kefu Chai <kchai@redhat.com>
ceph_volume_client.py's put_object_versioned() has only one test. Since
this only test is a negative test it may fail to assure that
put_object_versioned() works as expected with positive inputs even when
it completes successfully. Therefore, write a positive test for better
coverage.
Also, make sure the new test is both python 2 and python3 compatible.
Signed-off-by: Rishabh Dave <ridave@redhat.com>
test_put_object_versioned() is too generic for a negative test that
specifically tests if the version is verified before writing.
Signed-off-by: Rishabh Dave <ridave@redhat.com>
The test's success depends on whether CommandFailedError exception is
raised. This manner of testing is unreliable since there's no way to
check if the exception was raised by the statement (in the embedded
Python program) the test expects to.
This implies that this test's failure may go undetected. This is primary
reason why bugs like tracker #38946 occured despite of having a test.
Fixes: http://tracker.ceph.com/issues/39510
Signed-off-by: Rishabh Dave <ridave@redhat.com>
