[lustre-discuss] new mounted client shows lower disk space

Thomas Roth t.roth at gsi.de
Wed Nov 14 06:07:43 PST 2018


Hi,

your error messages are all well known - the one on the OSS will show up as soon as the Lustre modules
are loaded, provided you have some clients asking for the OSTs (and your MDT, which should be up by
then, is also looking for the OSTs).
The kiblnd_check_conns message I have also seen quite often, never with any OST problems.

Rather seems your OST take a lot of time to mount or to recover - did you check
/proc/fs/lustre/obdfilter/lustre-OST00*/recovery_status
?

Regards
Thomas

On 11/12/18 9:46 AM, fırat yılmaz wrote:
> Hi All
> OS=Redhat 7.4
> Lustre Version: Intel® Manager for Lustre* software 4.0.3.0
> 
> I have 72 osts over 6 oss with HA and 1 mdt serving to 195 clients over
> infiniband EDR.
> 
> After a reboot on client, lustre filesystem mounts on startup. It should be
> 2.1 TB area but lt starts with 350TB.
> 
> lfs osts command shows 90 percent of even numbered osts are ACTIVE and no
> information about other OSTs, as time passes like 1 hour or so, all OSTs
> become active and lustre area can be seen as 2.1 PB
> 
> 
> dmesg on lustre oss server:
> LustreError: 137-5: lustre-OST0009_UUID: not available for connect from
> 10.0.0.130 at o2ib (no target). If you are running an HA pair check that the
> target is mounted on the other server.
> 
> dmesg on client:
> LNet: 5419:0:(o2iblnd_cb.c:3192:kiblnd_check_conns()) Timed out tx for
> 10.0.0.5 at o2ib: 15 seconds
> Lustre: 5546:0:(client.c:2114:ptlrpc_expire_one_request()) @@@ Request sent
> has failed due to network error: [sent 1542009416/real 1542009426]
> req at ffff885f47610000 x1616909446641136/t0(0)
> o8->lustre-OST0030-osc-ffff885f75219800 at 10.0.0.8@o2ib:28/4 lens 520/544 e 0
> to 1 dl 1542009696 ref 1 fl Rpc:eXN/0/ffffffff rc 0/-1
> 
> I tested infiniband with ib_send_lat, ib_read_lat and no error occured
> I tested lnet ping with lctl ping 10.0.0.8 at o2ib , no error occured
> 12345-0 at lo
> 12345-10.51.22.8 at o2ib
> 
> Why some OST's  can be accesible while some are not in the same server?
> Best Regards.
> 
> 
> _______________________________________________
> lustre-discuss mailing list
> lustre-discuss at lists.lustre.org
> http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
> 




More information about the lustre-discuss mailing list