Bug 761815 (GLUSTER-83)

[Migrated from RT] - ticket 742 [http://support.gluster.com/rt/Ticket/Display.html?id=742]


Fri Jan 23 06:37:21 2009  	 guru - Ticket created  	 	 
version: glusterfs 2.0.0 qa10/871

[root@n2 guru]# ls
a bash dbench exn iozone kernel linux-2.6.24.4
linux-2.6.24.4.tar.bz2 rsync sort usr
[root@n2 guru]# rm -rf exn/
rm: cannot remove directory `exn/': Transport endpoint is not connected


* AFR over DHT and NUFA


* Out of the 4 servers, one server is down. b3-tcp and b3-ib would not
be accessible.


..
73: volume nufa
74: type cluster/nufa
75: option local-volume-name b1-ib
76: subvolumes b1-ib b1-tcp b2-ib b3-tcp
77: end-volume
78:
79: volume distribute
80: type cluster/distribute
81: subvolumes b2-tcp b3-ib b4-ib b4-tcp
82: end-volume
83:
84: volume replicate
85: type cluster/replicate
86: subvolumes nufa distribute
87: end-volume
..



However, once a directory has been created (along with the
subdirectories and files), why should removal of that directory fail?

--------------------------------------------------------------------------------
#   	Fri Jan 30 01:33:27 2009 	raghavendra - Comments added

Was the directory created after b3 was down?

On Fri Jan 23 06:37:21 2009, guru wrote:
> [root@n2 guru]# ls
> a bash dbench exn iozone kernel linux-2.6.24.4
> linux-2.6.24.4.tar.bz2 rsync sort usr
> [root@n2 guru]# rm -rf exn/
> rm: cannot remove directory `exn/': Transport endpoint is not connected
>
>
> * AFR over DHT and NUFA
>
>
> * Out of the 4 servers, one server is down. b3-tcp and b3-ib would not
> be accessible.
>
>
> ..
> 73: volume nufa
> 74: type cluster/nufa
> 75: option local-volume-name b1-ib
> 76: subvolumes b1-ib b1-tcp b2-ib b3-tcp
> 77: end-volume
> 78:
> 79: volume distribute
> 80: type cluster/distribute
> 81: subvolumes b2-tcp b3-ib b4-ib b4-tcp
> 82: end-volume
> 83:
> 84: volume replicate
> 85: type cluster/replicate
> 86: subvolumes nufa distribute
> 87: end-volume
> ..
>
>
>
> However, once a directory has been created (along with the
> subdirectories and files), why should removal of that directory fail?

the bug was seen with 2.0.0qa releases. Lot of fixes and improvements have happened both on distribute and replicate. I guess this issue is about stale linkfiles of distribute, but anyways, marking it "won't fix".