[Gluster-users] writev errors on our glusterFS

[mike d. kail]

i've had a glusterFS-2.01 server running on Amazon EC2 since June. 
instances are all ubuntu 8.04 with the libfuse2 pkg installed.  the FS
is mounted on 10 nodes and up until recently it had been working
flawlessly.  a few days ago we started getting strange write errors that
resulted in partial files being written to the glusterFS.  i did some
google'ing and then stopped the glusterfs process on all nodes and
removed the partial files from the underlying EXT3 filesystem. 
restarted and mounted everything and still got errors.  so, in hopes of
'solving' this, i upgraded to version 2.06, but we're still getting the
following errors on the server side.

x61e0c0: Invalid argument
[2009-08-27 10:15:41] D [server-protocol.c:1916:server_writev_cbk]
server: 33: WRITEV 1 (319495) ==> -1 (Invalid argument)


and now i'm also seeing these

[2009-08-27 13:32:47] D [server-protocol.c:3561:server_forget] brick:
22206: FORGET 16385 not found in inode table

anyone have any ideas on how this could have happened?  i'm just _very_
glad that it's not in production yet.

-- 
-------------------------------
mike d. kail
chief systems network architect
-------------------------------
cell: 858.922.2646
direct: 858.227.4801
mailto: mkail at attensity.com