RepoMirrors/ceph
1
0
Fork 0
mirror of https://github.com/ceph/ceph synced 2025-01-29 06:24:03 +00:00
Code Issues Packages Projects Releases Wiki Activity
9,201 Commits 216 Branches 394 Tags 1.2 GiB
03da8ba906
Commit Graph

9201 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Sage Weil
03da8ba906 osd: btrfs_snap by default 2010-02-09 08:09:37 -08:00
Sage Weil
48e3b00f42 osd: carry ondisk_read_lock if prepare_transaction may_read() 2010-02-09 08:09:23 -08:00
Sage Weil
153a37e7ac osd: order rmdir in remove_collection after deleting dir contents 2010-02-09 08:09:21 -08:00
Sage Weil
816ab37f52 filestore: fix old clobbering of commit_op_seq 
Update only at commit.  And add a newline.
 2010-02-09 08:09:19 -08:00
Sage Weil
408c81c210 osd: fix lock inversion with ondisk_write_unlock 
Drop write lock before taking pg lock.
 2010-02-09 08:09:16 -08:00
Sage Weil
ffb1b6c0b0 osd: wait for readable on primary after pulling an object 
Otherwise our attempt to push it back out again fail.

Note that currently this disables the previous push-right-after-
pull behavior.  We should either optimize the common case (full,
non-snapped object), or at least requeue the object so it's at
the front of the recovery queue so the fs cache will be hot.
 2010-02-05 16:38:41 -08:00
Sage Weil
aa8518879c mkcephfs: misc cephx fixes 2010-02-05 14:05:03 -08:00
Sage Weil
4d1899509c filestore: roll back to last clean snap on mount; clean up sanity checks 2010-02-05 13:14:08 -08:00
Sage Weil
9747e165b8 filestore: fix up commit_op_seq 
Note that we now _only_ update commit_op_seq when doing a proper commit.
That means that any additional commit the fs does will be unknown to us. If
we have a journal, that's fine, we'll replay it.  If not, we're screwed.

So if there's no journal, we need to be using btrfs_snap mode, and roll
back to the last snap.
 2010-02-05 11:32:48 -08:00
Sage Weil
e8783737ba filestore: use WorkQueue for ops (so we can have >1 thread issueing fs ops) 2010-02-05 11:23:42 -08:00
Sage Weil
31ad28e4ca Merge branch 'unstable' into filestore 
Conflicts:

	src/authtool.cc
	src/config.cc
 2010-02-05 10:14:51 -08:00
Sage Weil
496d764276 spinlock: make lockdep optional; off by default 
Spinlock tend to be inner locks; not much value in lockdep checking on
them.
 2010-02-05 09:42:00 -08:00
Sage Weil
147c6f51e3 osd: add uid to ceph_pg_pool; move encoding versions of rados.h 2010-02-04 16:18:43 -08:00
Sage Weil
213bd01002 monclient: return authetnication error 2010-02-04 14:35:21 -08:00
Sage Weil
035389a4c5 auth: tolerate bad authenticator replies (retry instead of crashing) 2010-02-04 14:24:23 -08:00
Sage Weil
91bf3e3cc9 mkcephfs: no scp -v 2010-02-04 13:49:42 -08:00
Sage Weil
dfc1b1115b auth: fix KeyStore interface 2010-02-04 13:48:23 -08:00
Sage Weil
53c0abc91b auth: simplify KeyRing map type, authmon add/importing 2010-02-04 11:03:27 -08:00
Sage Weil
7761ad97e6 cephx: separate KeyRing and RotatingKeyRing 
KeyRing stores keys (on disk).

RotatingKeyRing mediates access to the in-memory pool of rotating secrets
for a single service.
 2010-02-04 10:45:42 -08:00
Sage Weil
bee7be2f82 cephx: add struct_v throughout. some cleanup. 2010-02-04 09:31:30 -08:00
Sage Weil
ba515fe62a mkcephfs: generate cephx keys during mkfs 2010-02-03 16:40:16 -08:00
Sage Weil
40b90220c2 auth: use KeyRing encode/decode 2010-02-03 16:40:16 -08:00
Sage Weil
0b9f40ee7c todo 2010-02-03 16:40:16 -08:00
Yehuda Sadeh
329178d841 mount: set flags when getting -o sync 2010-02-03 15:55:19 -08:00
Sage Weil
6ea303014e mds: fix dumpcache 2010-02-03 12:50:41 -08:00
Sage Weil
6285b61d91 authtool: only create keyring if --create-keyring (or -c) 2010-02-03 11:58:40 -08:00
Sage Weil
f40957ebc3 config: rename 'keys file' to 'keyring' 2010-02-03 11:56:06 -08:00
Sage Weil
3ebf9a4920 filestore: optionally checkpoint with snaps 2010-02-03 11:36:52 -08:00
Sage Weil
5bdb348788 journal: make sure max_size is multiple of block_size 2010-02-03 11:31:30 -08:00
Sage Weil
54898b3222 mds: print setattr'd values with MClientRequest 2010-02-03 10:30:03 -08:00
Sage Weil
0fcd68b1d6 mds: only set smaller flag in setattr completion if actually truncating 
Otherwise we can erroneously trigger the truncation code when another
setattr overlaps with an in-progress truncate.
 2010-02-03 10:29:21 -08:00
Sage Weil
5ba3462805 authtool: only create keyring if --create-keyring (or -c) 2010-02-02 15:45:33 -08:00
Sage Weil
3a29b5c442 filestore: (try to) warn if not btrfs or ext3 
This can't distinguish between ext3 and ext4, still, and ext4 won't
work (because fsync(whatever) doesn't flush the whole journal).  Bleh.
 2010-02-02 15:28:19 -08:00
Sage Weil
6b09917ef9 vstart: writeahead journal by default 2010-02-02 15:12:04 -08:00
Sage Weil
7923a1929a filestore: adjust default sync intervals 2010-02-02 15:09:54 -08:00
Sage Weil
b8bf499ff8 filestore: warn on bad journaling, sync options 2010-02-02 15:08:07 -08:00
Sage Weil
c8f27eca73 journal: throttle journal 2010-02-02 15:08:04 -08:00
Sage Weil
c736e81d27 filestore: throttle op queue 2010-02-02 14:12:44 -08:00
Sage Weil
bd76ae934f todo 2010-02-02 11:52:40 -08:00
Sage Weil
37a45d52a9 osd: fix osd shutdown 2010-02-02 11:39:44 -08:00
Sage Weil
6963a6f697 journal: fix mis-merge 2010-02-02 11:18:49 -08:00
Sage Weil
c60edb4a3f Merge branch 'unstable' into filestore 
Conflicts:

	src/os/FileJournal.cc
 2010-02-02 11:05:46 -08:00
Sage Weil
24747836fe mds: dump_cache even when debug_mds < 2 
Make output more readable.
 2010-02-02 11:00:32 -08:00
Sage Weil
3642c2b229 journal: allow journal seq #'s to jump 
This happens when the regular store commits and we throw out now useless
journal entries.
 2010-02-02 11:00:32 -08:00
Sage Weil
f46b1b473f journal: make wrapping simpler 
Take out weirdness that tries to keep journal items contiguous.  No reason
not to split them across the end/beginning of the journal.  In the general
case, this is the same # of seeks because we have to rewrite the header
anyway.
 2010-02-02 11:00:32 -08:00
Yehuda Sadeh
99e316fbe1 osd: increase osdc protocol number 2010-02-01 16:14:05 -08:00
Yehuda Sadeh
e72cc233ec truncate: don't write beyong truncation with old trunc seq 
In a scenario where a truncation that followed a write got to
the osd before the preceding write, we shouldn't write beyond
that truncation when the write is handled in the osd.
 2010-02-01 16:14:05 -08:00
Yehuda Sadeh
9919b09540 truncate: initialize truncate_seq, truncate_size 2010-02-01 16:14:05 -08:00
Yehuda Sadeh
2f72778714 truncate: truncate information sent on each read/write op 
This fixes cases in which preceding client read/write ops arrive the
osd after a later truncation that arrived through the mds
 2010-02-01 16:14:05 -08:00
Sage Weil
af6803bd55 osd: quiet down about request perms 2010-02-01 14:53:41 -08:00