It's noting that complicated...
one mount one process that sticks around. I am using ib-verbs btw.
[root@RTPST205 ~]# ps -ef |grep
gluster
root 14423 13814 0 15:14 pts/1 00:00:00 grep gluster
[root@RTPST205 ~]# /usr/local/sbin/glusterfs -l
/var/glustersystemclient.log -f /etc/glusterfs-system.vol -d disable /system
[root@RTPST205 ~]# ps -ef |grep gluster
root 10972 1 0 12:16 ? 00:00:01
[glusterfs]
root 14438 13814 0 15:16 pts/1 00:00:00 grep gluster
[root@RTPST205 ~]# umount /system
[root@RTPST205 ~]# ps -ef |grep gluster
root 10972 1 0 12:16 ? 00:00:02
[glusterfs]
root 14451 13814 0 15:17 pts/1 00:00:00 grep gluster
-Mic
Sascha Ottolski wrote:
Am Freitag 29 Februar 2008 12:33:36 schrieb Anand Avati:
depends on whether you passed a mountpoint in the command line
argument. do you mean that the glusterfs client is still running
after you unmounted? or did you mean that before unmounting there
were two [glusterfs] and after unmounting there is just one (which
would be the server) ?
if I my add my 2 cents: as I reported in an earlier posting, it is
possible to mount the same mount-point several times, which results in
serveral glusterfs processes running. in such a case, you need to kill
all of them or umount several times.
of course, it would be best to prevent a second mount of a already
mounted gluster share (someone posted a recipe on how one could to
this; may this should be patched into the sources?).
Cheers, Sascha
avati
2008/2/29, Mickey Mazarick <mic@xxxxxxxxxxxxxxxxxx>:
Ah I see! I have an additional comment/question though.
I've noticed that when I unmount a gluster volume on a client the
[glusterfs] process is still running. Perhaps it thinks my cleint
spec is a server.
I have no instance of "type protocol/server" in the client spec.
What does it use to determine weather it's a client or a server?
-Mickey Mazarick
Anand Avati wrote:
Mickey,
in the latest codebase, there are no more server and client
programs. there is just one glusterfs (and glusterfsd is a
symlink to glusterfs). it behaves either as a client or server
according to the volume spec file given.
if the glusterfs program is passed a mountpoint it attaches a
fuse translator in a hardcoded way, so this preserves backward
compatibility. The new model also allows both protocol/server and
mount/fuse (mountability) in the same spec file which can be a
performance improvement in NUFA mode of operations.
avati
2008/2/29, Mickey Mazarick <mic@xxxxxxxxxxxxxxxxxx
<mailto:mic@xxxxxxxxxxxxxxxxxx>>:
This is very minor but it did break a kill script I wrote.
When I run the latest build, the server daemon shows up as:
[glusterfs]
in the process list.
Just an FYI :-)
Thanks!
-Mickey Mazarick
--
_______________________________________________
Gluster-devel mailing list
Gluster-devel@xxxxxxxxxx <mailto:Gluster-devel@xxxxxxxxxx>
http://lists.nongnu.org/mailman/listinfo/gluster-devel
--
If I traveled to the end of the rainbow
As Dame Fortune did intend,
Murphy would be there to tell me
The pot's at the other end.
--
_______________________________________________
Gluster-devel mailing list
Gluster-devel@xxxxxxxxxx
http://lists.nongnu.org/mailman/listinfo/gluster-devel
--
