Commit Graph

7926 Commits

Author SHA1 Message Date
Sage Weil
1b583a7ddf rev disk format 2009-09-03 10:52:31 -07:00
Sage Weil
5bc0dea9ff msgr todos 2009-09-03 10:39:07 -07:00
Sage Weil
68835958aa Merge branch 'unstable' into msgr 2009-09-02 16:31:08 -07:00
Sage Weil
a7fbab986b mds: use 0 for layout defaults, not -1
This simplifies callers that zero the structs.  And 0 is
clearly an illegal value, so it's safe to mean 'use the
default'.
2009-09-02 16:33:59 -07:00
Sage Weil
829ed7d772 Merge commit 'origin/unstable' into msgr
Conflicts:

	src/mon/ClientMonitor.cc
2009-09-02 16:13:14 -07:00
Sage Weil
85aa24228e msgr: fix entity_inst_t comparators
Do not use memcmp on unpacked type entity_inst_t!
2009-09-02 16:10:06 -07:00
Sage Weil
db494355e8 kclient: learn address during connect handshake 2009-09-02 16:10:06 -07:00
Sage Weil
218e4ed113 msgr: detect rank addr during connect handshake; clean up addr apis 2009-09-02 16:10:06 -07:00
Sage Weil
4330037a69 msgr: include my addr AND socket addr during connect handshake
This allows connecting side to detect their ip address.
2009-09-02 16:10:05 -07:00
Sage Weil
398d221e92 libceph: return 0, extern "C" 2009-09-02 15:51:24 -07:00
Greg Farnum
3b92ee70fa uclient/libceph: Framework for letting users specify file striping
and object sizes is in place.
2009-09-02 15:50:30 -07:00
Greg Farnum
0e812e6e41 mds: Handle the client requesting special stripe units, counts and object sizes. 2009-09-02 15:50:30 -07:00
Sage Weil
65f2d4a190 libceph: add ceph_version() 2009-09-02 13:57:19 -07:00
Sage Weil
a1d5533c43 debian: add lib{ceph,rados,crush}-dev packages 2009-09-02 13:55:31 -07:00
Sage Weil
1c1addc3a1 ceph: include version in ceph_fs.h 2009-09-02 13:35:23 -07:00
Sage Weil
5608be0ccb debian: fix lib* packages 2009-09-02 13:24:09 -07:00
Sage Weil
3e67dd8d0f kclient: whoami is __s64 2009-09-02 12:54:08 -07:00
Sage Weil
8d64e4d47b uclient: use client_t for client id 2009-09-02 12:53:59 -07:00
Sage Weil
a2c869c3ce mds: use client_t throughout 2009-09-02 12:53:47 -07:00
Sage Weil
27f6aa37ed msgr: 64bit entity name num 2009-09-02 11:32:21 -07:00
Sage Weil
7bb5895cd8 kclient: fix osd request kicking 2009-09-02 11:23:04 -07:00
Sage Weil
a790820a16 Makefile: force init.d script into /etc/init.d (not sysconfdir) 2009-09-02 10:49:55 -07:00
Sage Weil
d415a80782 man: fix mkcephfs man page 2009-09-02 10:49:52 -07:00
Sage Weil
f445219330 uclient: implement CREATE operation
This avoids an additional async cap wanted update in the usual
case.  And let's us pass in default layout info!

Untested...
2009-09-01 16:37:10 -07:00
Sage Weil
4ce269292b kclient: kill msgr->parent, dead msg pool code, prepare_pages arg 2009-09-01 16:16:53 -07:00
Sage Weil
5901023c16 kclient: osd restart fixes 2009-09-01 16:07:55 -07:00
Sage Weil
b7a9da1b9f kclient: set prepare_pages for osd client (der) 2009-09-01 16:02:42 -07:00
Sage Weil
738da1eec6 kclient: fix request kicking crash 2009-09-01 15:59:35 -07:00
Sage Weil
79b2740310 kclient: cleanup old FIXME comments 2009-09-01 15:51:29 -07:00
Sage Weil
ba2acd5244 kclient: improve osd error handling
We still r_resend even when we could fail, but it's just simpler
to do it that way.
2009-09-01 15:51:20 -07:00
Sage Weil
1ca917793f kclient: reuse osd struct when possible 2009-09-01 15:48:32 -07:00
Sage Weil
88c788f9d6 kclient: make mon hunt explicit; note in kern.log 2009-09-01 14:29:45 -07:00
Sage Weil
63aebffdfb mon: actually send updates to subscribers :) 2009-09-01 14:07:10 -07:00
Sage Weil
cdc4570f90 kclient: avoid get_random_int(), it's new
Just use get_random_bytes instead.
2009-09-01 14:02:39 -07:00
Sage Weil
ad0c75352e kclient: pick new mon on _fault_, not peer reset
The latter only occurs if we succeed in reconnecting and find that
the peer dropped the session (connect_seq didn't match).  That
won't happen if the mon is just down.
2009-09-01 13:08:28 -07:00
Sage Weil
6e0e406070 kclient: fix osdmap requests, subscription timeouts
Only request osdmap once, unless we restart our session.  Clean up
expired tests with helper.  Hunt if our monitor fails.
2009-09-01 12:34:28 -07:00
Sage Weil
eb5b6add84 kclient: no need for r_fwd_session
We always resend requests at the direction of the mds; the mds
won't forward for us.  (Otherwise the reply won't make it back!)
This simplifies forwards and retries.
2009-09-01 11:36:39 -07:00
Sage Weil
488cbb1b34 kclient: fix bad mds session put
From e5847ba5a9
2009-09-01 11:35:28 -07:00
Sage Weil
c31e9267d3 kclient: don't leak ref when queuing fault work 2009-09-01 11:04:17 -07:00
Sage Weil
ee2175fdb5 kclient: use subscribe to get mdsmap, osdmap updates 2009-09-01 11:04:03 -07:00
Sage Weil
2e49846903 msgr: print louder notice on bad banner; rev 2009-09-01 11:03:33 -07:00
Sage Weil
b2155c81b9 mon: subscribe duration in units of seconds, not ms 2009-09-01 11:03:15 -07:00
Sage Weil
1530e938d7 osd: remove ticket from MOSDOp 2009-09-01 10:52:27 -07:00
Greg Farnum
9063663005 uclient: fixing a backwards test in Client::put_cap_ref that resulted in
bad Inode ref counting.
2009-08-31 18:52:55 -07:00
Greg Farnum
d2e0186cc4 Hypertable: fix ups on the readdir interface that now uses getdnames. 2009-08-31 17:59:52 -07:00
Sage Weil
cdeb19e3e2 vstart: be quiet 2009-08-31 16:33:45 -07:00
Sage Weil
c47c926cf0 kclient: subscribe to mdsmap 2009-08-31 16:08:48 -07:00
Sage Weil
d863584a98 mon: clean up subscribe types a bits 2009-08-31 15:54:23 -07:00
Sage Weil
f7c1bec44a objecter: use one-time subscription request for osdmap 2009-08-31 15:54:05 -07:00
Sage Weil
0009d675c6 mon: do one-time subscriptions, too 2009-08-31 15:27:13 -07:00