I’m running lustre 1.8.1.1 on CentOS 5.4.
I ran the following:
mkfs.lustre --mgs --mdt /dev/hdb
mount -t lustre /dev/hdb /mnt/lustre
And everything worked without errors and I got the following messages in my logs:
Nov 25 11:23:58 localhost kernel: Lustre: lustre-MDT0000: new disk, initializing
Nov 25 11:23:58 localhost kernel: Lustre: MDT lustre-MDT0000 now serving lustre-
MDT0000_UUID (lustre-MDT0000/58b19442-2b9c-b7b6-4afc-c75a269effbc) with recovery enabled
Nov 25 11:23:58 localhost kernel: Lustre: 2792:0:(lproc_mds.c:271:lprocfs_wr_group_upcall()) lustre-MDT0000: group upcall set to /usr/sbin/l_getgroups
Nov 25 11:23:58 localhost kernel: Lustre: lustre-MDT0000.mdt: set parameter group_upcall=/usr/sbin/l_getgroups
Nov 25 11:23:58 localhost kernel: Lustre: Server lustre-MDT0000 on device /dev/hdb has started
And “df –h” shows the correct size of my lustre filesystem. However, if I try to ls (or access in other ways) to /mnt/lustre I get:
ls: /mnt/lustre: Not a directory
I was under the impression I didn’t need to create OSTs. Any ideas what I’m doing wrong? How do I access the lustre storage?
-Aaron
So this has started your meta-data service. You now need "data" service
(i.e. OST(s)).
> And “df –h” shows the correct size of my lustre filesystem. However,
> if I try to ls (or access in other ways) to /mnt/lustre I get:
>
>
>
> ls: /mnt/lustre: Not a directory
You cannot actually access the /mnt/lustre that you mounted on the MDS.
> I was under the impression I didn’t need to create OSTs.
You most certainly do. OSTs are where the data (i.e. the contents of
the files in your filesystem) are store. You also need to mount the
Lustre filesystem (as a whole) on a client and access it from that
client.
> Any ideas what I’m doing wrong? How do I access the lustre storage?
You create OSTs, mount them, then mount the filesystem on one or more
clients and access it from there.
It sounds like you really need to take a(nother) read through the
operations manual at manual.lustre.org to get a better picture of the
overall architecture.
b.
-Aaron
-----Original Message-----
From: lustre-disc...@lists.lustre.org
[mailto:lustre-disc...@lists.lustre.org] On Behalf Of Brian J.
Murrell
Sent: Wednesday, November 25, 2009 5:02 PM
To: lustre-...@lists.lustre.org
Subject: Re: [Lustre-discuss] getting "Not a directory" when accessing
mount point
On Wed, 2009-11-25 at 16:54 -0500, Aaron Siri wrote:
> I ran the following:
>
> mkfs.lustre --mgs --mdt /dev/hdb
>
> mount -t lustre /dev/hdb /mnt/lustre
So this has started your meta-data service. You now need "data" service
(i.e. OST(s)).
> And "df -h" shows the correct size of my lustre filesystem. However,
> if I try to ls (or access in other ways) to /mnt/lustre I get:
>
>
>
> ls: /mnt/lustre: Not a directory
You cannot actually access the /mnt/lustre that you mounted on the MDS.
> I was under the impression I didn't need to create OSTs.
You most certainly do. OSTs are where the data (i.e. the contents of
the files in your filesystem) are store. You also need to mount the
Lustre filesystem (as a whole) on a client and access it from that
client.
> Any ideas what I'm doing wrong? How do I access the lustre storage?
You create OSTs, mount them, then mount the filesystem on one or more
clients and access it from there.
It sounds like you really need to take a(nother) read through the
operations manual at manual.lustre.org to get a better picture of the
overall architecture.
b.
_______________________________________________
Lustre-discuss mailing list
Lustre-...@lists.lustre.org
http://lists.lustre.org/mailman/listinfo/lustre-discuss