Rohan, we are already aware of this bug and working on a fix. You can have a workaround for now by setting 'option flush-behind off' in write-behind translator. avati 2008/6/5 Rohan <rohan.thale@xxxxxxxxxxxxxxxx>: > Hi Avati, > > > > Occasionally we are getting following error on random clients. > > > > > > > > 2008-06-05 18:28:07 E [afr.c:2659:afr_flush] bricks: afrfdp is NULL, > returning EBADFD > > > > TLA Repo Revision: glusterfs--mainline--2.5--patch-770 > > Time : 2008-06-05 18:37:13 > > Signal Number : 11 > > > > /usr/local/sbin/glusterfs -s ACT-ST-013-R22.C2.GFS.dc1.in.com -l > /var/log/glusterfs/glusterfs-client-code-repository.log -L W > > ARNING /code-repository > > volume fuse > > type mount/fuse > > option direct-io-mode 1 > > option entry-timeout 1 > > option attr-timeout 1 > > option mount-point /code-repository > > subvolumes readonly-bricks > > end-volume > > > > volume readonly-bricks > > type features/filter > > subvolumes wb-bricks > > end-volume > > > > volume io-cache-bricks > > type performance/io-cache > > option page-size 2MB > > option cache-size 1024MB > > subvolumes wb-bricks > > end-volume > > > > volume wb-bricks > > type performance/write-behind > > option flush-behind on > > option aggregate-size 20MB > > subvolumes iothreads-bricks > > end-volume > > > > volume iothreads-bricks > > type performance/io-threads > > option cache-size 512MB > > option thread-count 400 > > subvolumes bricks > > end-volume > > > > volume afr-ns > > type cluster/afr > > option scheduler random > > subvolumes GFS-ACT-ST-001-NS GFS-ACT-ST-002-NS > > end-volume > > > > volume bricks > > type cluster/afr > > subvolumes GFS-ACT-ST-001-1 GFS-ACT-ST-002-1 > > end-volume > > > > volume GFS-ACT-ST-002-NS > > type protocol/client > > option transport-timeout 180 > > option remote-subvolume brick-ns > > option remote-port 6996 > > option remote-host act-st-014-r22.c2.gfs.dc1.in.com > > option transport-type tcp/client > > end-volume > > > > volume GFS-ACT-ST-001-NS > > type protocol/client > > option transport-timeout 180 > > option remote-subvolume brick-ns > > option remote-port 6996 > > option remote-host act-st-013-r22.c2.gfs.dc1.in.com > > option transport-type tcp/client > > end-volume > > > > volume GFS-ACT-ST-002-1 > > type protocol/client > > option transport-timeout 180 > > option remote-subvolume brick1 > > option remote-port 6996 > > option remote-host act-st-014-r22.c2.gfs.dc1.in.com > > option transport-type tcp/client > > end-volume > > > > volume GFS-ACT-ST-001-1 > > type protocol/client > > option transport-timeout 180 > > option remote-subvolume brick1 > > option remote-port 6996 > > option remote-host act-st-013-r22.c2.gfs.dc1.in.com > > option transport-type tcp/client > > end-volume > > > > frame : type(1) op(17) > > frame : type(0) op(0) > > frame : type(1) op(34) > > > > /lib64/libc.so.6[0x371d8300b0] > > /usr/local/lib/libglusterfs.so.0[0x2aaaaaab3fd2] > > /usr/local/lib/libglusterfs.so.0(dict_get+0x2e)[0x2aaaaaab407e] > > > /usr/local/lib/glusterfs/1.3.9/xlator/cluster/afr.so(afr_flush+0x67)[0x2aaaaaef1bd7] > > > /usr/local/lib/glusterfs/1.3.9/xlator/performance/io-threads.so[0x2aaaab1043a5] > > /usr/local/lib/libglusterfs.so.0(call_resume+0x78)[0x2aaaaaac07f8] > > > /usr/local/lib/glusterfs/1.3.9/xlator/performance/io-threads.so[0x2aaaab103a9d] > > /lib64/libpthread.so.0[0x371e4062f7] > > /lib64/libc.so.6(clone+0x6d)[0x371d8ce85d] > -- If I traveled to the end of the rainbow As Dame Fortune did intend, Murphy would be there to tell me The pot's at the other end.
