[Gluster-devel] 3.5.1qa4 performances

[Emmanuel Dreyfus]

On Thu, Dec 26, 2013 at 01:22:56AM -0500, Pranith Kumar Karampuri wrote:
>    I am not sure they are related. Could you check in both first and second bricks for the file?. If it is not present in both the bricks the file is already deleted and we can't figure out how it reached split-brain state :-(.

> > [afr-self-heal-common.c:233:afr_sh_print_split_brain_log]
> > 0-gfs351-replicate-0: Unable to self-heal contents of
> > '<gfid:5024cce3-a713-4bc4-acbc-129a4f2f0a66>' (possible split-brain).
> > Please delete the file from all but the preferred subvolume.- Pending
> > matrix:  [ [ 0 1 ] [ 2 0 ] ]

Weird, now I find them, while I could not inthe first place. Did I miss
them, or was it re-created afterward? (can it have the sam ehash?) Anyway:

Brick1 and brick 2 both have:
.glusterfs/50/24//5024cce3-a713-4bc4-acbc-129a4f2f0a66
Same trusted.gfid, 
brick1;trusted.afr.gfs351-client-0
   000   00 00 00 00 00 00 00 00 00 00 00 00                ............ 
brick1;trusted.afr.gfs351-client-1
   000   00 00 00 01 00 00 00 00 00 00 00 00                ............ 
brick2;trusted.afr.gfs351-client-0
   000   00 00 00 02 00 00 00 00 00 00 00 00   
brick2;trusted.afr.gfs351-client-1
   000   00 00 00 00 00 00 00 00 00 00 00 00                ............   

Brick 2 also have this, without any xattr and with size set to 0:
.glusterfs/indices/xattrop/5024cce3-a713-4bc4-acbc-129a4f2f0a66

Using inode number, I can find the culprit:
manu/usr/src/sys/arch/i386/compile/obj/GENERIC/midway.o

md5 show the file are the same, hence this is really a false positive.

-- 
Emmanuel Dreyfus
manu at netbsd.org