[Gluster-devel] mixed 3.7 and 3.6 environment

[Niels de Vos]

On Thu, Nov 12, 2015 at 05:11:32PM +0000, David Robinson wrote:
> Is there anyway to force a mount of a 3.6 server using a 3.7.6 FUSE client?
> My production machine is 3.6.6 and my test platform is 3.7.6.  I would like
> to test the 3.7.6 FUSE client but would need for this client to be able to
> mount both a 3.6.6 and a 3.7.6 server.

A 3.7.6 client will try to connect from an unprivileged port (> 1024).
Gluster servers with 3.6.x require connections from pivileged ports by
default (< 1024). The GETSPEC procedure that you see failing in the log,
is most likely because the GlusterD process terminates the connection
from the high port numbers.

It seems we've missed including release notes for the last 3.6.x
updates, but the "Known Issues" section of 3.6.3 still applies. It
contains the steps to allow connecting clients to use unprivileged
ports:

  https://github.com/gluster/glusterfs/blob/release-3.6/doc/release-notes/3.6.3.md#known-issues

Now, I expect you would like the 3.7.6 client to use privileged ports
instead. Unfortunately I am not sure if that is possible. I gave it a
quick try, but could not find a way to force the connection to GlusterD
to use a privileged port. Connecting to the bricks is possible like
that, when mounting with

  # mount -t glusterfs -o xlator-option=*.client-bind-insecure=no ...

This would then only require the change to /etc/glusterfs/glusterd.vol,
and not the server.allow-insecure option from the release notes.

HTH,
Niels


> 
> When I try to mount the 3.6.6 server using a 3.7.6 client, I get the
> following:
> 
> [root at ff01bkp glusterfs]# cat homegfs.log
> [2015-11-12 16:55:56.860663] I [MSGID: 100030] [glusterfsd.c:2318:main]
> 0-/usr/sbin/glusterfs: Started running /usr/sbin/glusterfs version 3.7.6
> (args: /usr/sbin/glusterfs --volfile-server=gfsib01a.corvidtec.com
> --volfile-server-transport=tcp --volfile-id=/homegfs.tcp /homegfs)
> [2015-11-12 16:55:56.868032] I [MSGID: 101190]
> [event-epoll.c:632:event_dispatch_epoll_worker] 0-epoll: Started thread with
> index 1
> [2015-11-12 16:55:56.871923] W [socket.c:588:__socket_rwv] 0-glusterfs:
> readv on 10.200.70.1:24007 failed (No data available)
> [2015-11-12 16:55:56.872236] E [rpc-clnt.c:362:saved_frames_unwind] (-->
> /lib64/libglusterfs.so.0(_gf_log_callingfn+0x192)[0x7f9e5507ea82] (-->
> /lib64/libgfrpc.so.0(saved_frames_unwind+0x1de)[0x7f9e54e49a3e] (-->
> /lib64/libgfrpc.so.0(saved_frames_destroy+0xe)[0x7f9e54e49b4e] (-->
> /lib64/libgfrpc.so.0(rpc_clnt_connection_cleanup+0x7a)[0x7f9e54e4b4da] (-->
> /lib64/libgfrpc.so.0(rpc_clnt_notify+0x88)[0x7f9e54e4bd08] )))))
> 0-glusterfs: forced unwinding frame type(GlusterFS Handshake) op(GETSPEC(2))
> called at 2015-11-12 16:55:56.871822 (xid=0x1)
> [2015-11-12 16:55:56.872254] E [glusterfsd-mgmt.c:1603:mgmt_getspec_cbk]
> 0-mgmt: failed to fetch volume file (key:/homegfs.tcp)
> [2015-11-12 16:55:56.872283] W [glusterfsd.c:1236:cleanup_and_exit]
> (-->/lib64/libgfrpc.so.0(saved_frames_unwind+0x205) [0x7f9e54e49a65]
> -->/usr/sbin/glusterfs(mgmt_getspec_cbk+0x490) [0x7f9e55556450]
> -->/usr/sbin/glusterfs(cleanup_and_exit+0x69) [0x7f9e555506d9] ) 0-:
> received signum (0), shutting down
> [2015-11-12 16:55:56.872299] I [fuse-bridge.c:5683:fini] 0-fuse: Unmounting
> '/homegfs'.
> [2015-11-12 16:55:56.872616] W [glusterfsd.c:1236:cleanup_and_exit]
> (-->/lib64/libpthread.so.0(+0x7df3) [0x7f9e53ee5df3]
> -->/usr/sbin/glusterfs(glusterfs_sigwaiter+0xe5) [0x7f9e55550855]
> -->/usr/sbin/glusterfs(cleanup_and_exit+0x69) [0x7f9e555506d9] ) 0-:
> received signum (15), shutting down
> David

> _______________________________________________
> Gluster-devel mailing list
> Gluster-devel at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-devel

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://www.gluster.org/pipermail/gluster-devel/attachments/20151113/66a1ae09/attachment.sig>