[Gluster-devel] add-brick crashes client

[Emmanuel Dreyfus]

On Fri, Aug 03, 2012 at 12:28:28PM +0000, Emmanuel Dreyfus wrote:
> > But nothing changes.

I found why: rebalance starts but fails on all servers. The rebalance
log show a sigsrgv, and gdb tells me it is at the same place as in 
the client:

#0  0xbbbc239c in synctask_wrap (old_task=0xb7f02400) at syncop.c:120120             task->ret = task->syncfn (task->opaque);
(gdb) print task
$1 = (struct synctask *) 0x0
(gdb) bt
#0  0xbbbc239c in synctask_wrap (old_task=0xb7f02400) at syncop.c:120
#1  0xbb8ed4e0 in swapcontext () from /lib/libc.so.12


I can apply the same patch on servers as I did on client (but it would
be nice if someone could tell me if it makes sense). But now that 
ls -lR has rebalanced the volumes, do I have a way to test again the
thing without starting over? I mean: is there a way to migrate data
out of a pair of bricks so that they can be removed?


-- 
Emmanuel Dreyfus
manu at netbsd.org