Yehuda Sadeh
5ae8e26cc8
class: fix adding of class library, when previously existed
2010-08-30 15:13:22 -07:00
Sage Weil
46a16b9ad2
mds: fix bad iterator usage in process_reconnected_caps()
...
We use p->second, so don't increment p++ early in the loop block.
2010-08-30 12:51:31 -07:00
Sage Weil
783d2ed0a4
mon: more useful class debug output
2010-08-28 07:34:04 -07:00
Greg Farnum
85814d45de
client: Make truncation work properly
...
The previous if block didn't work because inode->size was usually
changed well before handle_cap_trunc was ever invoked, so it never
did the truncation in the objectcacher! This was okay if you just truncated
a file and then closed it, but if you wrote a file, truncated part of it out,
and then wrote past the (new) end you would get reads that returned
the previously-truncated data out of what should have been a hole.
Now, we do the actual objectcacher truncation in update_inode_file_bits,
because all methods of truncation will move through there and this maintains
proper ordering.
2010-08-26 14:07:27 -07:00
Sage Weil
610b2e9dae
osd: fix class timeouts
2010-08-26 13:45:13 -07:00
Sage Weil
79f86c9b95
osd: clean up class loading code a bit
2010-08-26 13:45:13 -07:00
Sage Weil
20da3603d6
osdmap: whitespace
2010-08-26 12:33:14 -07:00
Sage Weil
9ee584201a
osd: use cluster msgr to share map with old hb peers
2010-08-26 12:33:11 -07:00
Sage Weil
114ed5aa2f
Merge branch 'testing' into unstable
2010-08-26 12:33:05 -07:00
Sage Weil
c3b97ffc1f
osd: mark down old _from connections too; and fix old _from debug output
2010-08-26 12:31:53 -07:00
Sage Weil
41cd1ea5b4
osd: don't mark down old _to peers that are still _from peers
...
It's harmless, but not ideal.
2010-08-26 12:31:53 -07:00
Sage Weil
d273c99d71
osd: always mark down old hb peers; send map update via cluster link
...
If we don't mark down the hb link immediately, we'll forget about it
because it won't be in the from or to set anymore, and if it does go down
later we'll end up with garbage in the logs.
Instead, always mark it down. Since we want to share our map with old
peers that are still up, do that via the cluster link instead, which is
reliably marked down if/when the peer goes down.
Signed-off-by: Sage Weil <sage@newdream.net>
2010-08-26 11:28:25 -07:00
Sage Weil
ae58862b31
Merge branch 'testing' into unstable
2010-08-26 10:00:37 -07:00
Sage Weil
7d1179e656
filejournal: clean up journal initialization sequence, error codes, output
...
EOPNOTSUPP was getting returned to the user without much explanation. This
should fix that!
2010-08-26 10:00:18 -07:00
Sage Weil
4e3003249c
mds: omit lock state in debug output when it's uninteresting (sync, no locks)
2010-08-25 21:36:19 -07:00
Sage Weil
6dae938316
qa: rename snapmove -> snaptest-parents
2010-08-25 21:36:19 -07:00
Sage Weil
f429dc8aaa
osd: don't reply on missing class
...
Instead, we queue ourselves in the waiting_for_class list. Broken by
f1eb9a8751
2010-08-25 16:24:41 -07:00
Sage Weil
c1a8c308d3
mds: only wrlock filelock during max_size changes on the head
2010-08-25 16:24:41 -07:00
Sage Weil
27cc3bdd8c
qa: add snaptest-dir-rename.sh
2010-08-25 16:24:41 -07:00
Greg Farnum
54d661aca6
qa: add blogbench test to qa
2010-08-25 16:10:57 -07:00
Sage Weil
49636759b0
debian: don't need to add .git_version in release script
2010-08-25 14:46:36 -07:00
Sage Weil
7fde620487
makefile: include .git_version in tarball
2010-08-25 14:45:55 -07:00
Sage Weil
b33c0511f9
mon: error out gracefully when we can't read the magic
2010-08-25 14:42:06 -07:00
Greg Farnum
655199a482
mds: fix invalid comparison.
...
We just want the code in this if block to execute if the previous if block did.
But the previous if block unlinked destdnl, so the comparison always fails! Use
a bool and set it appropriately to fix.
2010-08-25 12:06:08 -07:00
Sage Weil
98cf3c36cb
client: don't improperly conclude COMPLETE on empty dir
...
We can't trust the MDS values when we hold FILE_EXCL.
2010-08-24 16:31:19 -07:00
Sage Weil
1c176ee009
qa: clean up snaptest-* suite a bit
2010-08-24 16:01:47 -07:00
Greg Farnum
4f3407de99
client: don't print inode, it might be NULL.
...
This is especially bad because you can't start cfuse with debug level 20.
2010-08-24 15:11:56 -07:00
Greg Farnum
5b099f5c65
Revert "qa: remove random ordering from runallonce, for predictable resource matching"
...
This snuck in by mistake in the leaks merge.
This reverts commit 8778bacf3b
2010-08-24 13:55:30 -07:00
Sage Weil
f07fdeee6a
client: only includee xattrs in caps msg if we are flushing that bit
2010-08-24 13:06:13 -07:00
Sage Weil
99674a5fae
client: queue capsnap on any dirty metadata
2010-08-24 13:05:48 -07:00
Sage Weil
520706df3c
Merge branch 'unstable' of ssh://ceph.newdream.net/home/sage/ceph.newdream.net/git/ceph into unstable
2010-08-24 12:58:00 -07:00
Sage Weil
bc8ee67cc7
Merge branch 'rootsnap' into unstable
...
Conflicts:
src/mds/Server.cc
2010-08-24 12:17:57 -07:00
Sage Weil
d34c7789e1
mds: fix cow_old_inode to properly save xattrs
2010-08-24 12:15:05 -07:00
Sage Weil
c9b1c54ad5
mds: fix snaprealm create split
...
Need to check if diri->snaprealm is set _before_ popping the dirty inode
(and new snaprealm!).
2010-08-24 12:14:59 -07:00
Sage Weil
ec02fe38a7
mds: no snaps in system dirs, but root is ok
2010-08-24 12:14:46 -07:00
Sage Weil
2f2cccb4ca
mds: drop old debug output
2010-08-24 12:14:41 -07:00
Sage Weil
f32a880bc2
mds: rename project_past_parent -> project_past_snaprealm_parent
2010-08-24 12:14:38 -07:00
Sage Weil
ac12b0cb78
mds: drop unused projected_srnode list
2010-08-24 12:14:35 -07:00
Sage Weil
7172a637b7
qa: snaptest-xattrwb.sh
2010-08-24 12:13:53 -07:00
Sage Weil
10ed52012b
qa: chmod +x snaptest-authwb.sh
2010-08-24 12:13:53 -07:00
Sage Weil
08920f4e81
mds: expose projected values for all locks to loner (not just filelock)
...
There is nothing special about filelock in this case. If the client is
the loner, we should use the projected values. This is important when
we are looking at a snapid on the head (see snaptest-authwb.sh).
2010-08-24 12:13:53 -07:00
Sage Weil
25cc1451c8
mds: do null snapflush as soon as we have all file wr caps
...
The snapflush only gets delayed on the client by data writeback, so we can
do the null snapflush as soon as we have those bits.
2010-08-24 12:13:53 -07:00
Sage Weil
cd92072e71
mds: fix null deref of cap in do_cap_update when doing intermediate snap updates
2010-08-24 12:13:53 -07:00
Greg Farnum
c4ab621ed6
mon: add clock sync backoff to handle_lease_ack
2010-08-24 11:54:49 -07:00
Greg Farnum
6998372a8f
mon: Adjust lease time checks.
...
For some reason this fixes (apparently incorrect) warnings I saw
while running a 2-mon system on my dev server.
2010-08-24 11:49:43 -07:00
Greg Farnum
09fbe8650e
Merge branch 'leaks' into unstable
...
Conflicts:
src/mds/Locker.cc
2010-08-24 10:35:35 -07:00
Greg Farnum
6ace230b8f
Merge branch 'unstable' into leaks
...
Conflicts:
src/mds/Locker.cc
2010-08-23 23:41:41 -07:00
Sage Weil
1d8770be27
Merge branch 'auth_change_waiters' into unstable
2010-08-23 20:31:17 -07:00
Sage Weil
067d8bb0d1
mds: don't auth_pin non-auth wrlock targets
2010-08-23 20:23:14 -07:00
Sage Weil
8d462435a7
mds: don't auth_pin non-auth rdlock targets
...
This generates lots of lock traffic that is unneeded with the authchange
notification machinery.
2010-08-23 20:23:14 -07:00
