Lustre discuss - Nov 2007 - heartbeat + drbd +lustre

Dear All,

Could somebody give me a howto to make this work? I want to setup
(replace an old one) an mds based on CentOS 5. I want to setup
heartbeat v2 (crm) + drbd 0.8 + lustre 1.6.3.

My problem is I don''t find any useful howto related to 
heartbet v2 + crm + drbd .

They write everywhere, it''s the same as in v1, but put crm yes to
ha.cf, and use gui or edit the right xml file. It''s OK, but how?

I''m sorry for the stupid question.

Thank you very much,

tamas

On Tue, Nov 20, 2007 at 09:48:34PM +0100, Papp Tamas wrote:

Well, this was not so much help, maybe about this:

When I try to mount the filesistem (still lustre), I see in the log:

Nov 30 17:20:22 teszt1 crmd: [2768]: info: do_lrm_rsc_op: Performing
op=mount-mdt_start_0 key=34:4:cc234ee0-2545-474b-8ea3-dd61d5de47f2)
Nov 30 17:20:22 teszt1 Filesystem[3526]: INFO: Running start for
/dev/drbd0 on /mdt
Nov 30 17:20:22 teszt1 kernel: Lustre: OBD class driver,
info at clusterfs.com
Nov 30 17:20:22 teszt1 kernel:         Lustre Version: 1.6.3
Nov 30 17:20:22 teszt1 kernel:         Build Version:
1.6.3-19691231190000-PRISTINE-.cache.build.BUILD.lustre-kernel-2.6.18.lustre.linux-2.6.18-8.1.14.el5_lustre.1.6.3smp
Nov 30 17:20:22 teszt1 kernel: Lustre: Added LNI 192.168.0.66 at tcp
[8/256]
Nov 30 17:20:22 teszt1 kernel: Lustre: Accept secure, port 988
Nov 30 17:20:22 teszt1 kernel: Lustre: Lustre Client File System;
info at clusterfs.com
Nov 30 17:20:22 teszt1 Filesystem[3526]: INFO: Starting filesystem
check on /dev/drbd0
Nov 30 17:20:22 teszt1 lrmd: [2765]: info: RA output:
(mount-mdt:start:stdout) fsck 1.40.2.cfs1 (12-Jul-2007)
Nov 30 17:20:22 teszt1 lrmd: [2765]: info: RA output:
(mount-mdt:start:stderr) fsck: fsck.lustre: not found
Nov 30 17:20:22 teszt1 lrmd: [2765]: info: RA output:
(mount-mdt:start:stderr) fsck: Error 2 while executing fsck.lustre for
/dev/drbd0
Nov 30 17:20:23 teszt1 kernel: kjournald starting.  Commit interval 5
seconds
Nov 30 17:20:23 teszt1 kernel: LDISKFS FS on drbd0, internal journal
Nov 30 17:20:23 teszt1 kernel: LDISKFS-fs: mounted filesystem with
ordered data mode.
Nov 30 17:20:23 teszt1 kernel: SELinux: initialized (dev drbd0, type
ldiskfs), not configured for labeling
Nov 30 17:20:23 teszt1 kernel: LustreError:
3633:0:(obd_mount.c:1231:server_kernel_mount()) premount
/dev/drbd0:0x0 ldiskfs failed: -13, ldiskfs2 failed: -19.  Is the
ldiskfs module available?
Nov 30 17:20:23 teszt1 kernel: LustreError:
3633:0:(obd_mount.c:1536:server_fill_super()) Unable to mount device
/dev/drbd0: -13
Nov 30 17:20:23 teszt1 kernel: LustreError:
3633:0:(obd_mount.c:1927:lustre_fill_super()) Unable to mount  (-13)
Nov 30 17:20:23 teszt1 lrmd: [2765]: info: RA output:
(mount-mdt:start:stderr) mount.lustre: mount /dev/drbd0 at /mdt
failed: Permission denied
Nov 30 17:20:23 teszt1 Filesystem[3526]: ERROR: Couldn''t mount
filesystem /dev/drbd0 on /mdt
Nov 30 17:20:23 teszt1 crmd: [2768]: ERROR: process_lrm_event: LRM
operation mount-mdt_start_0 (call=14, rc=1) Error unknown error
Nov 30 17:20:23 teszt1 crmd: [2768]: info: build_operation_update:
Digest for 4:1;34:4:cc234ee0-2545-474b-8ea3-dd61d5de47f2
(mount-mdt_start_0) was 5e579999da68f8d23996cdfcde038242
Nov 30 17:20:23 teszt1 crmd: [2768]: info: log_data_element:
build_operation_update: digest:source <parameters fstype="lustre"
directory="/mdt" target_role="started"
device="/dev/drbd0"/>
Nov 30 17:20:24 teszt1 crmd: [2768]: info: do_lrm_rsc_op: Performing
op=mount-mdt_stop_0 key=1:5:cc234ee0-2545-474b-8ea3-dd61d5de47f2)
Nov 30 17:20:24 teszt1 Filesystem[3659]: INFO: Running stop for
/dev/drbd0 on /mdt
Nov 30 17:20:24 teszt1 crmd: [2768]: info: process_lrm_event: LRM
operation mount-mdt_stop_0 (call=15, rc=0) complete




But it''s working if the fstype is ext3. I''m out of ideas.
Please
somebody give me a push:)

It''s working from command line:

mount -t lustre /dev/drbd0 /mdt


This is the relevant xml:

<primitive class="ocf" provider="heartbeat"
type="Filesystem"
id="mount-mdt">
  <instance_attributes id="ia_mount-mdt">
    <attributes>
      <nvpair id="mount-mdt_fstype" name="fstype"
value="lustre"/>
      <nvpair id="mount-mdt_mountpoint" name="directory"
value="/mdt"/>
      <nvpair id="mount-mdt_device" name="device"
value="/dev/drbd0"/>
    </attributes>
  </instance_attributes>
</primitive>


Thank you very much,

tamas

On Nov 30, 2007  18:06 +0100, Papp Tamas wrote:
> (mount-mdt:start:stderr) fsck: Error 2 while executing fsck.lustre for
> /dev/drbd0
You need to make the "fsck pass" number in your /etc/fstab be 0
(skip).
It isn''t safe to run e2fsck on a shared device except as part of the
failover itself.  If you are doing this ONLY as part of the failover,
then you can link e2fsck to fsck.lustre (using Lustre-patched e2fsprogs
only).
> Nov 30 17:20:23 teszt1 crmd: [2768]: ERROR: process_lrm_event: LRM
> operation mount-mdt_start_0 (call=14, rc=1) Error unknown error
> Nov 30 17:20:23 teszt1 crmd: [2768]: info: build_operation_update:
> Digest for 4:1;34:4:cc234ee0-2545-474b-8ea3-dd61d5de47f2
> (mount-mdt_start_0) was 5e579999da68f8d23996cdfcde038242
> Nov 30 17:20:23 teszt1 crmd: [2768]: info: log_data_element:
> build_operation_update: digest:source <parameters
fstype="lustre"
> directory="/mdt" target_role="started"
device="/dev/drbd0"/>
> Nov 30 17:20:24 teszt1 crmd: [2768]: info: do_lrm_rsc_op: Performing
> op=mount-mdt_stop_0 key=1:5:cc234ee0-2545-474b-8ea3-dd61d5de47f2)
> Nov 30 17:20:24 teszt1 Filesystem[3659]: INFO: Running stop for
> /dev/drbd0 on /mdt
> Nov 30 17:20:24 teszt1 crmd: [2768]: info: process_lrm_event: LRM
> operation mount-mdt_stop_0 (call=15, rc=0) complete
> 
> 
> 
> 
> But it''s working if the fstype is ext3. I''m out of ideas.
Please
> somebody give me a push:)
> 
> It''s working from command line:
> 
> mount -t lustre /dev/drbd0 /mdt
> 
> 
> This is the relevant xml:
> 
> <primitive class="ocf" provider="heartbeat"
type="Filesystem"
> id="mount-mdt">
>   <instance_attributes id="ia_mount-mdt">
>     <attributes>
>       <nvpair id="mount-mdt_fstype" name="fstype"
value="lustre"/>
>       <nvpair id="mount-mdt_mountpoint"
name="directory"
> value="/mdt"/>
>       <nvpair id="mount-mdt_device" name="device"
value="/dev/drbd0"/>
>     </attributes>
>   </instance_attributes>
> </primitive>
Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.