[Lustre-discuss] Can't re-activate osc after crash (on 1.8.0)

[Jakob Goldbach]

Hi,

A OSS crashed last night - I deactivated on mds and clients by using 

lctl --device <no> deactivate

After I got the OSS server up, I tried to mount OST by got this:

[  404.081109] LDISKFS FS on cciss/c0d1, internal journal
[  404.081533] LDISKFS-fs: mounted filesystem with ordered data mode.
[  404.081910] LDISKFS-fs: file extents enabled
[  404.093549] LDISKFS-fs: mballoc enabled
[  404.117445] Lustre: MGC172.16.14.10 at tcp: Reactivating import
[  418.180860] LustreError: 137-5: UUID 'backup-OST0012_UUID' is not
available  for connect (no target)
[  418.181563] LustreError:
2933:0:(ldlm_lib.c:1826:target_send_reply_msg()) @@@ processing error
(-19)  req at ffff810077a4f400 x1304450061357696/t0 o8-><?>@<?>:0/0 lens
368/0 e 0 to 0 dl 1245189724 ref 1 fl Interpret:/0/0 rc -19/0

(last lines repeats).

After re-activation on MDS it tries to connect to the OSS:

[1217955.946590] Lustre:
4781:0:(import.c:508:import_select_connection()) Skipped 8 previous
similar messages
[1218161.548655] Lustre: Request x1304459003964424 sent from
backup-OST0012-osc to NID 172.16.14.38 at tcp 56s ago has timed out (limit
56s).

So it seems I have a chicken-and-egg problem. OST wont't mount, MDS
can't connect.

Any ideas?

BTW, Vanilla 2.6.22.19 with lustre 1.8.0 (both build by me).

Thanks,
/Jakob