This is unlikely to be noticed by anybody, but it is a big change. Document
in the PendingReleaseNotes and bump up the librados minor version number
to 68.
Signed-off-by: Greg Farnum <greg@inktank.com>
The error message helpfully references the -m and -c CLI options for
specifying monitors, but this code can be invoked from non-core librados
client applications so that's unfortunately not kosher. Remove the
reference.
Fixes#5979.
Signed-off-by: Greg Farnum <greg@inktank.com>
Not sure why but this seems to resolve a linking problem when loading
classes:
2013-08-17 13:28:19.015776 7fb2bcffa700 0 _load_class could not open class /usr/lib/rados-classes/libcls_hello.so (dlopen failed): /usr/lib/rados-classes/libcls_hello.so: undefined symbol: cls_log
2013-08-17 13:28:19.015786 7fb2bcffa700 -1 osd.4 12 class hello open got (5) Input/output error
In any case, it's simpler.
Signed-off-by: Sage Weil <sage@inktank.com>
Previously we would create an empty ESubtreeMap when we opened the log
segment and then immediately journal a second one that created the root
and mdsdir. More importantly, for the second ESubtreeMap, we would not
wait for it to commit before requesting the ACTIVE state, leading to
#4894.
Instead, break start_new_segment() into two steps: one that creates the
in-memory LogSegment tracking structure, and one that journals the
ESubtreeMap. Open things early and write the (one) ESubtreeMap at the
end of boot_create().. and then wait for it.
Fixes: #4894
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Yan, Zheng <zheng.z.yan@intel.com>
This appears to be one source of confusion for new users that leads to
a failure to form an initial mon quorum. See comments on
http://tracker.ceph.com/issues/4924
Signed-off-by: Sage Weil <sage@inktank.com>
Use sftp to upload to a directory that only this user and ceph devs can
access.
Distribute an ssh key to connect to the account. This will let us revoke
the key in the future if we feel the need. Also distribute a known_hosts
file so that users have some confidence that they are connecting to the
real ceph drop account and not some third party.
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Dan Mick <dan.mick@inktank.com>
We don't want to clone_range from clones too many times.
For now, just skip the cloning if there are too many holes.
Fixes: #5985
Signed-off-by: Samuel Just <sam.just@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
These aren't used by the C++ code at all, but in order for
rados_conf_get to find them, they need to be listed. They're
consumed by ceph_rest_api.
Signed-off-by: Dan Mick <dan.mick@inktank.com>
cls/hello: hello, world rados class
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Reviewed-by: Samuel Just <sam.just@inktank.com>
Reviewed-by: Loic Dachary <loic@dachary.com>
We do not try to merge rx buffers currently. Make that explicit and
documented in the code that it is not supported. (Otherwise the
last_read_tid values will get lost and read results won't get applied
to the cache properly.)
Signed-off-by: Sage Weil <sage@inktank.com>
Consider a sequence like:
1- start read on 100~200
100~200 state rx
2- truncate to 200
100~100 state rx
3- start read on 200~200
100~100 state rx
200~200 state rx
4- get 100~200 read result
Currently this makes us crash on
osdc/ObjectCacher.cc: 738: FAILED assert(bh->length() <= start+(loff_t)length-opos)
when processing the second 200~200 bufferhead (it is too big). The
larger issue, though, is that we should not be looking at this data at
all; it has been truncated away.
Fix this by marking each rx buffer with the read request that is sent to
fill it, and only fill it from that read request. Then the first reply
will fill the first 100~100 extend but not touch the other extent; the
second read will do that.
Signed-off-by: Sage Weil <sage@inktank.com>
Class methods are marked with RD and WR to help the OSD decide when we need
to flush objects or require certain permissions. Ensure that methods do
not step outside their advertised capabilities by keeping a counter of rd
and wr ops we perform in do_osd_ops() and making sure that class methods,
and any ops the indirectly call, do not break the rules.
Signed-off-by: Sage Weil <sage@inktank.com>
This method is problematic because it both writes/mutates and returns data,
which means that an untimely client disconnect or peering event will result
in a success to the client with no payload.
It has not been used since v0.52 (18054ba46f)
which is pre-bobtail; so this change breaks compatibility with pre-bobtail
librbd clients (at least for image creation).
Signed-off-by: Sage Weil <sage@inktank.com>
We were somewhat inadvertantly returning a data payload for write
operations. This was a side-effect of the OpContext::ops field being a
reference to MOSDOp::ops: the return data would end up there, and then
the MOSDOpReply ctor would copy it.
Fix this by breaking the ref, and making the do_op() logic also claim
return result data for error values (so that errors can return data to the
caller).
Signed-off-by: Sage Weil <sage@inktank.com>
common/Preforker.h: In member function 'void Preforker::daemonize()':
common/Preforker.h:97:40: warning: ignoring return value of 'ssize_t write(int, const void*, size_t)', declared with attribute warn_unused_result [-Wunused-result]
Signed-off-by: Sage Weil <sage@inktank.com>
Fix by restructuring code to hoist common code and have only one
place where admin_socket is actually called.
Signed-off-by: Dan Mick <dan.mick@inktank.com>
'osd crush set' should only be used to update already existing items on
the map whereas 'osd crush add' should be able to 'add and update' items.
Considering at that point we are effectively adding a new item to the
crush map, use 'add' instead of 'set'.
Signed-off-by: Joao Eduardo Luis <joao.luis@inktank.com>
