On 09/14/2009 01:21 PM, David Saez Padros wrote:
Hi
Is there any problem in having both the gluster file system
mounted in the traditional way and booster using the same
'mount point' ?
Due to some system calls not being handling by Gluster - I think I would
suggest it.
It's not perfect - but at least any calls that fall through will still
be handled properly. For example, if an application calls fopen(), which
is not on the GlusterFS list of overridden system calls the last time I
checked, then at least the fopen() will be intercepted by FUSE rather
than fail altogether.
One poster (not sure if it was you) suggested that they access the local
volume directly, rather than through a server. I saw this as being a
potential problem, since the feature/locks was not the same for both
local and remote accesses. I could see something similar if the
feature/locks was being handled separately for FUSE and booster. Not
having tried it myself - on theory alone - I suggest both talk to a
local server, allowing the server to schedule concurrent accesses,
rather than both trying to schedule concurrent accesses to the
underlying storage/posix as separate processes trying to operate
independently.
Cheers,
mark
--
Mark Mielke<m...@mielke.cc>
_______________________________________________
Gluster-users mailing list
Gluster-users@gluster.org
http://gluster.org/cgi-bin/mailman/listinfo/gluster-users