Re: [Gluster-devel] is umount supposed to work?

[Amar Tumballi]

Emmanuel,

Replies inline.

I am still working on glusterfs/NetBSD, and it is getting close to something
actually usable. I still have a few issues, one of which is unmounting.

Thats a good news.

 

When I use umount(8), I send a FUSE_DESTROY to glusterfs. That has no
effect at all. Looking at the source, I can see that fuse_destroy()
just returns succes and do nothing.

Is it on purpose? What about adding a kill(getpid(), SIGTERM) here?
What is the prefered way of unmounting right now?

IMO its not on purpose. below patch should be better option than calling a kill().

-----

diff --git a/xlators/mount/fuse/src/fuse-bridge.c b/xlators/mount/fuse/src/fuse-bridge.c

index dafc0a9..902c68a 100644

--- a/xlators/mount/fuse/src/fuse-bridge.c

+++ b/xlators/mount/fuse/src/fuse-bridge.c

@@ -2922,6 +2922,9 @@ fuse_destroy (xlator_t *this, fuse_in_header_t *finh, void *msg)

         send_fuse_err (this, finh, 0);

 

         GF_FREE (finh);

+

+        if (this)

+                this->fini (this);

 }

 

-----

 

Another problem: invoking glusterfs with -s and without --volfile-id
leads to a SIGSEGV because volfile_id is NULL. This can be fixed by
adding an error message if --volfile-id is supplied without -s. Another
approach would be to let the administrator supply the volfile_id in the
-s string, for instance with glusterfs -s address@hidden /mnt
(I suggest this syntax rather than gfs.example.net:gfs1 just in case one
want to add support for specifying an alternate port some day).

Filed a bug here:  http://bugs.gluster.com/show_bug.cgi?id=2934

Patch sent to fix it: http://patches.gluster.com/patch/7232/

With these things should be smooth for you.

Regards,

Amar