1
0
mirror of git://sourceware.org/git/lvm2.git synced 2024-12-22 17:35:59 +03:00
Commit Graph

26 Commits

Author SHA1 Message Date
Patrick Caulfield
e7f55bb4c0 Get clvmd to use libcman rather than cman ioctl calls. This makes
it forward-compatible with the new userland CMAN in cluster head.

To build it you will need the libcman header & library installed.
2006-03-14 14:18:34 +00:00
Patrick Caulfield
4bb7474e8a Don't send a signal to kill threads that are idling nicely as it upsets them.
This seems to cure bz#159727 on SMP systems.

Alasdair, can you include this patch in the lvm2-cluster package please ?
2005-08-09 10:39:57 +00:00
Patrick Caulfield
470b46cee1 Log an error if clvmd can't resolve a host name got from CCS
Fix potential spin loop in clvmd
2005-06-14 10:35:02 +00:00
Patrick Caulfield
5a3f4a794d Don't defer closing of FDs in clvmd as it can cause trouble. 2005-06-13 10:16:21 +00:00
Patrick Caulfield
30bda7761e Tidy clvmd's SIGHUP handler so it doesn't do all that work. 2005-04-19 10:36:42 +00:00
Patrick Caulfield
2b7e4013cd Make clvmd work around some "limitations" in gulm's node state notifications.
Also make clvmd debuglog timestamps a little more helpful.
2005-04-13 13:50:07 +00:00
Patrick Caulfield
11270d07bf Don't allow user operations to start until the lvm thread is fully up.
Hopefully finally nails bz#146056
2005-04-01 13:01:01 +00:00
Alasdair Kergon
779b7713cf Tighten signal handlers. 2005-03-21 14:16:16 +00:00
Patrick Caulfield
a6c23d8b8b Don't take out the lvm_thread_lock at startup - it only protects the jobs list.
DEBUGLOG() message now print threadid rather than PID which is more useful.
2005-03-09 14:08:11 +00:00
Patrick Caulfield
3329bbfdf3 make clvmd FDs close-on-exec, to avoid warnings when running lvm via popen.
clvmd-gulm unlocks VG & orphan locks at startup in case they are stale.
clvmd-gulm now unlocks VG & orphan locks if client dies.
2005-03-07 17:03:44 +00:00
Patrick Caulfield
c9808c329d Cope with more than one message arriving at the TCP socket, also
fix some instances where the length in the message was wrong (cman
code didn't notice this because it is packet-based comms anyway)
2005-02-18 15:31:32 +00:00
Patrick Caulfield
ceda2b1426 Fix thread shutdown race which could cause clvmd to hang in pthread_join. 2005-02-08 09:05:58 +00:00
Patrick Caulfield
c9f9a8b2b3 Make clvmd announce it's startup and cluster connection in syslog. 2005-01-24 15:31:13 +00:00
Patrick Caulfield
6f4d25ec7f Fix clvmd startup bug introduced in cman/gulm amalgamation. bz#145729
Improve reporting of node-specific locking errors so you'll get
somthing a little more helpfiul than "host is down" - it will now tell
you /which/ host it thinks is down.
2005-01-21 11:35:24 +00:00
Patrick Caulfield
5c7fc7c4f7 You can now build clvmd with cman & gulm support in the same binary.
./configure --with-clvmd
wil do this by default. Or you can choose which you want with
./configure --with-clvmd=gulm    or
./configure --with-clvmd=cman

When clvmd with both included is run, it will automatically detect the cluster
manager in use.
2005-01-13 13:24:02 +00:00
Patrick Caulfield
b9fb9b206e Fix off-by-one error in cluster_locking that could case read hangs. 2005-01-07 14:22:49 +00:00
Patrick Caulfield
0d19475588 Improve clvmd failure message if it's already running.
Allow user to kill clvmd during initialisation.
2005-01-05 14:41:54 +00:00
Alasdair Kergon
a266258f4c Configure/makefile tidy. 2004-11-26 18:07:17 +00:00
Patrick Caulfield
bb7b7c6d5f Tidy the socket callbacks so that all the work is done outside the
main loop.
2004-11-16 10:55:01 +00:00
Patrick Caulfield
badd6fb7ba Report detailed errors in cluster initialisation to syslog, and point the user
at syslog if clvmd fails to intialise.
2004-11-11 14:51:23 +00:00
Patrick Caulfield
3d4159b6bf Make clvmd -V display the lvm version number too rather than just
the protocol version (which is not that useful and doesn't change very often).
2004-10-06 12:36:47 +00:00
Patrick Caulfield
bfd4b01407 Revert the fork back to where it was as it seems to confuse pthreads.
Instead, the parent now waits for the daemon to signal that it has
completed successfully (or not) so it can return status to the user.
2004-10-06 10:12:34 +00:00
Patrick Caulfield
896318f475 Fork a little later in the startup sequence so that we can return
an error code if the cluster infrastructure isn't there.
2004-09-30 14:16:28 +00:00
Patrick Caulfield
8ef0b7b9ef Keep client locks (VG locks usually) in their own hash table so
we can actually have more then one of them per client.
2004-09-24 09:39:57 +00:00
Patrick Caulfield
985f52fe34 Make the thread handling a little less cavalier. In particular, calling
pthread_exit in a signal handler was a /really/ bad idea.
2004-09-23 12:51:56 +00:00
Alasdair Kergon
d3c8211fef Add cluster support. 2004-06-24 08:02:38 +00:00