Unmap takes a device now, and uses the major number
to find the right sequence number in sysfs.
Also updated the man page.
Signed-off-by: Josh Durgin <josh.durgin@dreamhost.com>
Also include missing headers and make it linkable with C,
since it's duplicated from mount.ceph.
Signed-off-by: Josh Durgin <josh.durgin@dreamhost.com>
testlibrbd.c: In function 'write_test_data':
testlibrbd.c:191: warning: comparison between signed and unsigned integer expressions
testlibrbd.c: In function 'aio_read_test_data':
testlibrbd.c:207: warning: comparison between signed and unsigned integer expressions
testlibrbd.c: In function 'read_test_data':
testlibrbd.c:222: warning: comparison between signed and unsigned integer expressions
Signed-off-by: Sage Weil <sage@newdream.net>
We accidentally broke it in our last round of changes to support
standby-replay; patch it up again.
Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com>
testlibrbd.c: In function 'write_test_data':
testlibrbd.c:191: warning: comparison between signed and unsigned integer expressions
testlibrbd.c: In function 'aio_read_test_data':
testlibrbd.c:207: warning: comparison between signed and unsigned integer expressions
testlibrbd.c: In function 'read_test_data':
testlibrbd.c:222: warning: comparison between signed and unsigned integer expressions
Signed-off-by: Sage Weil <sage@newdream.net>
Apparently using "as" in "except" blocks is a new (and not
backwards-compatible) thing in Python 2.6.x. Who knew?
Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
We now register observers on mon_osd_full_ratio and nearfull_ratio
to update the OSD full/nearfull lists.
Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com>
Previously these sets were only modified when applying an incremental,
which might have caused problems when restarting monitors. Whoops!
Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com>
Objects can now register as configuration observers interested in a
subset of the configuration keys. The observers will be told exactly
which keys have changed.
The first user is dout, which now no longer needs the infamous SIGHUP
hack to know when to reopen the config file.
librados: Remove rados_reopen_log, which was basically a means for the
library user to trigger the SIGHUP behavior.
Changes are accumulated and applied all at once by apply_changes. This
function is called as part of common_init, and after every call to
injectargs.
Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
Since logging options are per-config, logically DoutStreambuf instances
should also be per-config. This also allows us to eliminate the
"if (uninitialized)" checks at the beginning of every call to dout.
Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
This was used to implement log_to_stdout, but that doesn't exist any
more. Instead of we have log_to_stderr and the other stuff.
Signed-off-by: Colin McCabe <colin.mccabe@dreamhost.com>
pg_remove has been included for longer than we've had versions
in the struct, so this check for end is useless -- if pg_remove
wasn't encoded we're already broken by decoding the version at
the beginning.
Signed-off-by: Greg Farnum <gregory.farnum@dreamhost.com>
lenny doesn't have a suitable libgoogle-perftools-dev, and
release.sh edits it out of build-deps. Detect that and tell
configure that not having tcmalloc is ok.
This should make 05c281bfa9
unnecessary.
Signed-off-by: Tommi Virtanen <tommi.virtanen@dreamhost.com>
This is not strictly needed as of 05c281bfa9,
but that reverting is hopefully only temporary.
Without this, with 05c281 undone, non-mainstream architectures
would fail to build.
Signed-off-by: Tommi Virtanen <tommi.virtanen@dreamhost.com>
This is now the admin's job. Removes a lot of code with limited testing
and coverage.
We rev the internal monitor protocol because the state machine ids changed.
This should not affect the on-disk format. Just stop and restart all the
monitors at once during the upgrade.
Signed-off-by: Sage Weil <sage.weil@dreamhost.com>
Simplify the class loading code to just load out of a local directory.
Do not request classes from the monitor.
Signed-off-by: Sage Weil <sage.weil@dreamhost.com>
These are added to the LogSegment list on the slaves, but also need to be
removed from that list when we replay a COMMIT|ROLLBACK or when the op's
fate is determined during the resolve stage.
This fixes a crash like
./include/elist.h: In function 'elist<T>::item::~item() [with T =
MDSlaveUpdate*]', in thread '0x7fb2004d5700'
./include/elist.h: 39: FAILED assert(!is_on_list())
ceph version 0.26 (commit:9981ff90968398da43c63106694d661f5e3d07d5)
1: (MDSlaveUpdate::~MDSlaveUpdate()+0x59) [0x4d9fe9]
2: (ESlaveUpdate::replay(MDS*)+0x422) [0x4d2772]
3: (MDLog::_replay_thread()+0xb90) [0x67f850]
4: (MDLog::ReplayThread::entry()+0xd) [0x4b89ed]
5: (()+0x7971) [0x7fb20564a971]
6: (clone()+0x6d) [0x7fb2042e692d]
ceph version 0.26 (commit:9981ff90968398da43c63106694d661f5e3d07d5)
1: (MDSlaveUpdate::~MDSlaveUpdate()+0x59) [0x4d9fe9]
2: (ESlaveUpdate::replay(MDS*)+0x422) [0x4d2772]
3: (MDLog::_replay_thread()+0xb90) [0x67f850]
4: (MDLog::ReplayThread::entry()+0xd) [0x4b89ed]
5: (()+0x7971) [0x7fb20564a971]
Fixes: #1019
Signed-off-by: Sage Weil <sage@newdream.net>
