[Gluster-devel] Segfault

[Kevan Benson]

Yep, the config loads but is non-functional with your change.  I assumed 
there was a good chance it wouldn't work, but it would have allowed for 
a better performance profile and easier file administration of a simple 
HA 2 server cluster if it did, so I figured it was worth a try.  Even if 
it served files fine, I figured there might be large lock problems to 
overcome.  Anyways, thanks for the pointer on the namespace collisions.

Krishna Srinivas wrote:
> Hi Kevan,
>
> A volume can not have two parents. Both a-ds and b-ds have
> afr1 and afr2 as parents. You could use different names for
> afr2's children, though the connection going to the same storage
> as a-ds b-ds. This is because translators store data in data
> structures shared across the translators using the volume name
> as the key. So this key has to be unique.
>
> If unify sees more than one file amongst its children, it returns
> EIO for  that particular file. Let us know how it behaves though.
>
> Regards
> Krishna
>
> On 10/17/07, Kevan Benson <kbenson at a-1networks.com> wrote:
>   
>> Another test configuration, another segfault.  ;)  Admittedly, I'd doing
>> some odd stuff, like dual afr membership like so on the client:
>> afr-ns = a-ns, b-ns
>> afr1 = a-ds, b-ds
>> afr1 = b-ds, a-ds
>> unify = afr1, afr2 / ns = afr-ns
>>
>> Servers simply export a namespace (ns) and a posix-locked share (ds).
>>
>> Clients and servers start fine, any access from the client causes both
>> servers to segfault as below.
>>
>> Yes, I'm aware this may not be a supported configuration, I just figured
>> I'd try, and then see how locking worked (as I can see there might be a
>> posix locking problem with this if it doesn't segfault).
>>
>> fuse glfs4
>> glusterfs tla 522
>>
>> backtrace:
>> #0  0x00a37546 in pthread_spin_lock () from /lib/tls/libpthread.so.0
>> #1  0x005a6853 in fd_destroy (fd=0xb7e00d40) at xlator.c:282
>> #2  0x0011f35a in server_closedir_cbk (frame=0xb7e00b68,
>> cookie=0xb7e005b8, this=0x9ae5088, op_ret=0, op_errno=17) at
>> server-protocol.c:976
>> #3  0x005a8389 in default_closedir_cbk (frame=0xaaaaaaaa,
>> cookie=0xb7e004b0, this=0x9ae4a88, op_ret=0, op_errno=17) at defaults.c:986
>> #4  0x00112cd4 in posix_closedir (frame=0xb7e004b0, this=0x9ae44f8,
>> fd=0xb7e00d40) at posix.c:287
>> #5  0x005a8415 in default_closedir (frame=0xb7e005b8, this=0x9ae4a88,
>> fd=0xb7e00d40) at defaults.c:997
>> #6  0x00123684 in server_closedir (frame=0xb7e00b68, bound_xl=0x9ae4a88,
>> params=0xb7e005f0) at server-protocol.c:3968
>> #7  0x00126b13 in notify (this=0x9ae5088, event=2, data=0x9b1daa8) at
>> server-protocol.c:5637
>> #8  0x005aa7d8 in transport_notify (this=0x9b1daa8, event=1) at
>> transport.c:154
>> #9  0x005aafa1 in sys_epoll_iteration (ctx=0xbff7ef20) at epoll.c:53
>> #10 0x005aaac5 in poll_iteration (ctx=0xbff7ef20) at transport.c:300
>> #11 0x080494ce in main (argc=5, argv=0xbff7eff4) at glusterfsd.c:319
>>
>>
>> --
>>
>> -Kevan Benson
>> -A-1 Networks
>>
>>
>> _______________________________________________
>> Gluster-devel mailing list
>> Gluster-devel at nongnu.org
>> http://lists.nongnu.org/mailman/listinfo/gluster-devel
>>
>>     
>
>
> _______________________________________________
> Gluster-devel mailing list
> Gluster-devel at nongnu.org
> http://lists.nongnu.org/mailman/listinfo/gluster-devel
> .
>
>   


-- 

-Kevan Benson
-A-1 Networks