[Lustre-discuss] Error when using lfs

[Andreas Dilger]

On Nov 29, 2007  15:06 -0600, Jeremy Mann wrote:
> I configured a new cluster today with 1.6.0.1 and I'm having a strange
> problem. I'm trying to set the stripe size, but every time I do, lfs hangs
> and the follow error is logged in /var/log/kern:
> 
> LustreError: 26571:0:(lib-move.c:95:lnet_try_match_md()) Matching packet
> from 12345-0 at lo, match 26080 length 856 too big: 744 left, 744 allowed
> Lustre: 26571:0:(lib-move.c:1644:lnet_parse_put()) Dropping PUT from
> 12345-0 at lo portal 10 match 26080 offset 0 length 856: 2
> 
> The headnode is the MDT and MGS and I have 17 OSTs. After a few minutes,
> it times out and the logs read:
> 
> LustreError: 9853:0:(mdc_locks.c:414:mdc_enqueue()) ldlm_cli_enqueue: -4
> Lustre: 26563:0:(ldlm_lib.c:497:target_handle_reconnect())
> laredofs-MDT0000: d89ea9bb-4dd5-e0ae-d718-5841d1555a20 reconnecting
> Lustre: laredofs-MDT0000-mdc-0000010033e8a800: Connection restored to
> service laredofs-MDT0000 using nid 0 at lo.

This looks a known problem with directory default striping and ACLs active
at the same time.  Bug 13358.

Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.