Sage Weil [Fri, 15 Jun 2012 17:42:49 +0000 (10:42 -0700)]
cls_rbd: check for LAYERING feature for parent methods
You need to set the feature on the image before you can use these methods.
We *could* allow get_parent, but that just means they'll get ENOENT instead
of ENOEXEC, and the latter is more informative.
Samuel Just [Tue, 19 Jun 2012 16:11:57 +0000 (09:11 -0700)]
PG: improve find_best_info
07f853db3982e68b952a337cf91cbf7ec0709de9 is actually too conservative,
it suffices to find any info with a last_update of at least the least
last_update from the last period to go active. An info from a previous
interval is acceptable if the last interval never reported a commited
operation and thus still has the same last_update.
Signed-off-by: Samuel Just <sam.just@inktank.com> Reviewed-by: Sage Weil <sage@inktank.com>
Sage Weil [Mon, 18 Jun 2012 21:00:06 +0000 (14:00 -0700)]
mon: gracefully handle slow 'ceph -w' clients
If we are sending log updates to a client (ceph -w), and they are far
enough behind to drop behind first_committed, include a friendly message
in their stream but continue.
Drop useless return value from _create_sub_incremental(). Assert that we
can read the state file.
Samuel Just [Sat, 16 Jun 2012 00:09:42 +0000 (17:09 -0700)]
PG: best_info must have a last_epoch_started as high as any other info
We disregard incomplete infos during find_best_info, but we can't an
info with a last_epoch_started less that of the incomplete info.
This should avoid cases like #2462. In that case, it appears that
a peer with empty info/log was chosen as authoritative even though
there was a non-empty incomplete peer.
Samuel Just [Sat, 16 Jun 2012 00:09:42 +0000 (17:09 -0700)]
PG: best_info must have a last_epoch_started as high as any other info
We disregard incomplete infos during find_best_info, but we can't an
info with a last_epoch_started less that of the incomplete info.
This should avoid cases like #2462. In that case, it appears that
a peer with empty info/log was chosen as authoritative even though
there was a non-empty incomplete peer.
Greg Farnum [Sat, 16 Jun 2012 00:11:02 +0000 (17:11 -0700)]
cls_rbd: add failure logging to all callers of read_key.
Previously, read_key was printing out error messages when it failed.
But because some callers are fine with it failing on ENOENT, and they
all have more context anyway, it no longer does so and they need
their own logging output. Now they have it.
Greg Farnum [Tue, 12 Jun 2012 20:03:36 +0000 (13:03 -0700)]
cls_rbd: add locking functions.
Users can now place shared or exclusive locks, can remove their own
locks, can break the locks of others, and can list locks. All these
functions are considered PUBLIC_EXEC to match the other rbd methods.
Yehuda Sadeh [Tue, 12 Jun 2012 21:42:03 +0000 (14:42 -0700)]
rgw: obj copy respects -metadata-directive
Fixes #2542. The old behavior just merged src object attrs
and provided attributes. The new (and correct) behavior looks
at the x-[amz|rgw|...]-metadata-directive and either copies
the source attrs, or replaces them with the provided attrs.
Greg Farnum [Mon, 11 Jun 2012 17:19:54 +0000 (10:19 -0700)]
librbd: only clear the needs_refresh flag on successful updates.
To support this, add a refresh_seq counter which is incremented every
time we get a notify. This way we can handle the concern from 7add136f907876a20aa2a68297c66b8259afd9a7, in which we get a notify
in the middle of doing a refresh.
Sage Weil [Thu, 14 Jun 2012 00:15:33 +0000 (17:15 -0700)]
osdcap: move semantics of cap check into OSDCap class
The class needs to understand the semantics of possible actions in enough
detail to apply the capabilities it describes, so it makes sense to pass
all of that information into a check method here. That is preferable to
exposing the internal structure of the capability itself to an external
function to apply the same logic.
Specifically, this fixes the problem where we return the first matching
OSDCapSpec, not a union of all matching specs, and thus fixes a problem
where applying a cap like 'allow r, allow pool foo rwx' to a write
operation in pool foo.
Sage Weil [Wed, 13 Jun 2012 20:42:24 +0000 (13:42 -0700)]
osdcap: drop unused 'auid' arg from get_cap(), is_match()
The intent was to do a cap like 'allow pools-that-i-own ...', but in
practice you just do 'allow pool <myuid> ...'.. the actual auid you
authenticate as is redundant, since the cap itself is authenticated.
Sage Weil [Wed, 13 Jun 2012 18:05:43 +0000 (11:05 -0700)]
Makefile: link gtest statically
The problem:
- the unittests link against gtest, and gtest is not installed. that's
normally fine, but...
- rbd and rados api unit tests link against gtest, and are installed
by 'make install'. they are needed for teuthology runs, etc.
- if we build gtest as an .la library, we can only control whether *all*
or *no* .la libraries are linked statically.
- we want librados to be linked dynamically.
The solution:
- build gtest as .a instead of a libtool library
- link it statically, always.
Unit test binaries are bigger now. Oh well...
Fixes: #2331 Signed-off-by: Sage Weil <sage@inktank.com>
Sage Weil [Wed, 13 Jun 2012 02:16:28 +0000 (19:16 -0700)]
osd: reimplement OSDCap struct
The old OSDCaps had a few undesireable features:
- deny
- weird 'default permissions' semantics when auid matches and no other
cap is specified
- difficult to understand, maintain parser
This version is renamed and has a few nice features:
- spirit grammar for parsing
- simple internal grant structure with simple semantics
- support for object prefix maches
- partial support for per-class (not yet implmeented by the OSD)
Samuel Just [Tue, 12 Jun 2012 19:53:02 +0000 (12:53 -0700)]
PG: track purged pgs during active
See bug #2462.
The following sequence could cause a log assuming a non-empty pg
to an empty replica:
1. primary sends query to stray
2. stray sends notify to primary
3. primary sends purge to stray removing stray from peer_info
4. stray recieves query and sends a notify
5. stray recieves purge and purges its pg
6. primary recieves notify from stray and adds it to peer_info
note: peer_info[stray] is now wrong
7. acting set changes, primary is still primary, stray is replica
8. primary sends log to replica based on incorrect info from 6.
This patch adds a purged_peer set which is populated during purge_strays
and cleared during start_peering_interval. The primary will ignore
notifies from the peer once the peer is in this set.
projects / ceph.git / log