[Lustre-discuss] MDT mount options (after upgrade 1.8.7 to 2.4.3)

Gretchen Zwart zwart at astro.umass.edu
Wed Jun 18 12:29:59 PDT 2014


Hi,

Thanks for that information. I figured out mount options for the OSTs but am now having trouble mounting on a client.

Did you do any renaming of OSTs (e.g. lctl --replace-nids)? Did you make any changes at all related to naming scheme? As I recall you said you had 1.8.9wc clients as well as some 1.8.6...

I see the following in the log files:
     MDT using 1.8 OSC name scheme
     For 1.8 interoperability, rename obd type from mds to mdt

For the client it is complaining about existing device:
     Device lustre-OST0000-osc-ffff88022517bc00 already exists at 3, won't add

I don't think you mentioned if you set up a second MDT. I'm currently working off a copied MDT and may go back to using drbd/ha MDT. It was quite oversized...

I am grateful to you for sharing your experience.
Regards,
Gretchen


On 06/18/2014 11:28 AM, Mohr Jr, Richard Frank (Rick Mohr) wrote:
>
> On Jun 17, 2014, at 9:46 AM, Gretchen Zwart <zwart at astro.umass.edu> wrote:
>
>> I now need to sort out correct mount options for combined MDT/MGS (was
>> w/ failover and quotas). I think the failover.node is the bad option
>> but am not sure of the correct syntax. Presumably I will have to fix
>> the params with tunefs.lustre for the OSTs as well...
> You will need to use tunefs.lustre with the --servicenode=<NID> option to specify failover hosts.
>
>>    Read previous values:
>> Target:     lustre-MDT0000
>> Index:      0
>> Lustre FS:  lustre
>> Mount type: ldiskfs
>> Flags:      0x105
>>               (MDT MGS writeconf )
>> Persistent mount opts: errors=remount-ro,iopen_nopriv,user_xattr
>> Parameters: failover.node=10.16.10.231 at o2ib,10.0.11.231 at tcp
>> mdt.quota_type=ug2 mdt.group_upcall=/usr/sbin/l_getgroups
> Those mount options match exactly what I have on my MDT, so I am not sure what Lustre is complaining about.
>
> One other thing I forgot to mention before was that we had to set vm.zone_reclaim_mode=0 on our MDT to avoid problems with lustre threads getting stuck for long periods of time.  (This isn't directly related to the mount issue you currently have, but I thought I would mention it since you might find it important later on.)
>

-- 
Gretchen Zwart
UMass Astronomy Dept. 619E Lederle
710 North Pleasant ST
Amherst,MA 01003
(413) 577-2108




More information about the lustre-discuss mailing list