Sage Weil
a08d9619d5
rados: move librados.h, rados_bencher.h
...
Also include rados_bencher.h in Makefile.am.
2009-12-04 10:03:49 -08:00
Sage Weil
dc24351187
mds: avoid scatter writebehind from predirty_*
...
predirty_journal_parents does wrlock_start with
nowait=true. Make sure we don't then simple_lock if
lock->is_updated(), as that will call scatter_writebehind,
which will barf if a log event is currently open (which
it is in this case).
2009-12-03 16:27:06 -08:00
Sage Weil
e71e3619fc
crush: fix whitespace
2009-12-03 16:27:06 -08:00
Sage Weil
f7008441b3
logger: s64, not long
...
So we can log larger numbers.
2009-12-03 16:27:06 -08:00
Sage Weil
a4257d8782
buffer: optional buffer debug output
...
This interferes with the dout crap because we can't #include it, so you
need to turn all that output off to avoid getting jumbled output. Or
better yet make sure you don't daemon() but still log dout to a file.
Fixed up the output parser to find leaks.
2009-12-03 16:21:10 -08:00
Sage Weil
25e8202f69
ceph: fix message leaks
2009-12-03 16:21:10 -08:00
Sage Weil
f3c83bfe9f
init-ceph: optionally copy executable to remote host before running it
...
This both preserves the version of the binary that we ran (for interpreting
core files) and avoids any possibility of SIGBUS due to nfs lameness when
rarely used text pages get faulted back in.
2009-12-03 16:21:10 -08:00
Greg Farnum
7fd1a2171a
mds: typos *sigh*
2009-12-03 13:51:08 -08:00
Greg Farnum
e2939a1199
mds: Cleanup/doc updates
2009-12-03 13:44:48 -08:00
Greg Farnum
a010652ff8
mds: Add export killpoints
2009-12-03 13:44:48 -08:00
Greg Farnum
b6dba80ada
mds: Add killpoints for mds export
2009-12-03 13:43:43 -08:00
Greg Farnum
00dbe2bfd9
mds: Add mds_kill_export_at asserts.
...
Create and begin documenting doc/killpoints.txt
2009-12-03 13:43:43 -08:00
Greg Farnum
e439bd33a0
config: add kill arguments for mds import/export
2009-12-03 13:43:42 -08:00
Sage Weil
b709a72833
mds: bracket mds journal events with {start,submit}_entry
...
This lets us catch nested journal events when they happen,
instead of down the line when dir/inode version numbers
look weird.
2009-12-03 12:12:55 -08:00
Sage Weil
95ee211343
todo
...
snaprealm errors on kclient when ls -al .ceph/mds0/stray
2009-12-02 16:57:07 -08:00
Sage Weil
adbd7d8b87
dropped old aleung mds branch code
...
See aleung_mds_security branch.
2009-12-02 16:56:07 -08:00
Sage Weil
714a9af30d
mon: fix allocation of low global_ids after mon restart
...
We need to set last_allocated in init(). Also avoid the
-1 casting and just use 0 instead.
2009-12-02 15:03:50 -08:00
Sage Weil
89603b6bca
test_ioctls: print preferred_osd value from GET_LAYOUT ioctl
2009-12-02 14:46:13 -08:00
Sage Weil
54b8537c75
hash: fix whitespace
2009-12-02 12:03:53 -08:00
Sage Weil
fd0195a9d8
mds: set mdr->in[n] in rdlock_path_xlock_dentry
...
This keeps things consistent with what we'd get with
path_traverse. Namely, it means that if done_locking gets
set to true, then a future call to rdlock_path_pin_ref can
bail out and return mdr->in[n] and get a correct value.
This fixes
touch mnt/a
ls mnt
touch mnt/.ceph
mds hang on the create request.
2009-12-02 12:00:19 -08:00
Sage Weil
378e476923
mds: do not include /.ceph in readdir results
2009-12-02 11:26:24 -08:00
Greg Farnum
7928120495
librados: set id in messenger
2009-12-01 17:32:38 -08:00
Greg Farnum
364e045df8
rados: read benchmark is asynchronous.
2009-12-01 17:32:38 -08:00
Greg Farnum
8efea57667
Change hump-backed variable names.
...
Initialize data->in_flight!
2009-12-01 17:32:38 -08:00
Greg Farnum
18b909a544
rados: pull sequential benchmark into its own function
2009-12-01 17:32:38 -08:00
Greg Farnum
3d43376988
rados: remove iTime from object names for complete determinism
2009-12-01 17:32:38 -08:00
Greg Farnum
bcbafa6d18
rados: Handle errors slightly better
2009-12-01 17:32:38 -08:00
Greg Farnum
6a8bda51de
rados: split up function into more subfunctions
2009-12-01 17:32:38 -08:00
Greg Farnum
b79063658c
rados: Deterministically fill data, reverse count:time order
2009-12-01 17:32:38 -08:00
Greg Farnum
15373402bc
librados: minor comments
2009-12-01 17:32:38 -08:00
Greg Farnum
ae9edc8c06
rados: Move variables into data structure; rename to be generic
2009-12-01 17:32:38 -08:00
Greg Farnum
e328d28918
rados: Move benchmark functions into separate file
2009-12-01 17:32:38 -08:00
Greg Farnum
6250b9329f
qa: Rename a file for future clarity.
2009-12-01 17:32:38 -08:00
Sage Weil
83e2b9cc79
qa: specify test path as argument ot runallone.sh
2009-12-01 16:18:56 -08:00
Sage Weil
9b5b0f9a47
config: preserve argv[0] if appropriate
...
This makes 'cfuse' show up in 'df' output.
2009-12-01 16:19:19 -08:00
Sage Weil
defeedb545
qa: test mdstable failures
2009-12-01 16:19:19 -08:00
Sage Weil
c280aace53
todo
...
the lockdep was probably due to the leak of request_mutex?
2009-12-01 16:19:19 -08:00
Sage Weil
8bc7d2262f
mds: inject mds failure during mdstable operations
2009-12-01 15:39:22 -08:00
Sage Weil
5bd7404856
mds: log successful client reconnect
2009-12-01 15:00:56 -08:00
Sage Weil
77d26c9edb
sepia: asdf
2009-12-01 15:00:56 -08:00
Sage Weil
89beb68e0e
mds: journal/replay rollback on MDSTableServer
...
We journal prepare and commit; need to do rollback to!
2009-12-01 15:00:56 -08:00
Sage Weil
6c34e67190
mds: track _note_prepared/committed in normal path, too
...
This was broken by cd6f3817
.
2009-12-01 15:00:56 -08:00
Sage Weil
5c0be5d28c
mds: pass do_eval through to adjsut_subtreeauth
2009-12-01 15:00:56 -08:00
Sage Weil
8dd0991ea5
vstart: add [mds.a] section when no -x
2009-12-01 12:07:07 -08:00
Sage Weil
75d4ce8908
mds: drop requests on closed|closing|etc sessions
2009-12-01 10:38:47 -08:00
Sage Weil
beb2fb9588
ceph.spec: mostly in line with .deb
...
No packages for radosgw or libcrush/ceph/rados, though.
2009-11-25 21:45:11 -08:00
Sage Weil
4c94d162b9
object: add missing include
...
For sprintf().
2009-11-25 21:34:51 -08:00
Sage Weil
99379f607b
mds: ignore reconnect if session state is closed
...
On reconnect, if we didn't have a session from the journal, our Session
state will be 'closed'. Act accordingly.
2009-11-25 21:23:57 -08:00
Sage Weil
d66f2903bc
todo
2009-11-24 19:02:06 -08:00
Sage Weil
05bd3f712f
mds: don't renew caps (or touch_session) on closed session
...
Also assert that we're on the state list before touching.
2009-11-24 19:02:06 -08:00