Phil Schwan
2006-May-19 07:36 UTC
[Lustre-discuss] 0-config with multiple file systems having problems
On Wed, 2004-04-28 at 18:41, Andrew_Norrie@veritasdgc.com wrote:> > Using v1.0.4, pre-patched kernel, RH8.0. > I''ve successfully configured the 0-config setup with 1 filesystem, but > cannot get a file system to mount when 2 or more file systems are > configured. > Are 2 or more file systems possible with 0-config or is there something > wrong with the following configuration?What you have looks correct, but I would suggest trying something slightly different, because I don''t think the multi-FS-plus-zeroconf case has been exceptionally well-tested.> [root@node0 lustre]# cat config.sh > --add node --node node0 > --add net --node node0 --nid node0 --nettype tcp > --add mds --node node0 --mds mds1 --fstype ext3 --dev /tmp/mds1 --size > 100000 > --add lov --lov lov1 --mds mds1 --stripe_sz 1048576 --stripe_cnt 1 > --stripe_pattern 0 > --add node --node node10 > --add net --node node10 --nid node10 --nettype tcp > --add ost --node node10 --lov lov1 --fstype ext3 --dev /scratch/ost1 --size > 100000 > --add node --node client > --add net --node client --nid client --nettype tcp > --add mtpt --node client --mds mds1 --lov lov1 --path /mnt/lustre1 > --add mds --node node0 --mds mds2 --fstype ext3 --dev /tmp/mds2 --size > 100000 > --add lov --lov lov2 --mds mds2 --stripe_sz 1048576 --stripe_cnt 1 > --stripe_pattern 0 > --add ost --node node10 --lov lov2 --fstype ext3 --dev /scratch/ost2 --size > 100000 > --add mtpt --node client --mds mds2 --lov lov2 --path /mnt/lustre2What if you remove this last line, and add: --add node --node client2 --add net --node client2 --nid ''*'' --nettype tcp --add mtpt --node client2 --mds mds2 --lov lov2 --path /mnt/lustre2 and then mount fs1 with: node0:/mds1/client and fs2 with: node0:/mds2/client2 Does that help? -Phil
Andrew_Norrie@veritasdgc.com
2006-May-19 07:36 UTC
[Lustre-discuss] 0-config with multiple file systems having problems
Yes, having a separate "client profile" for each file system did the trick. Thanks ... Andrew --------------------------------------------------------- Andrew Norrie, Veritas GeoServices Ltd. 2200, 715 5 Ave. S.W. Calgary, Alberta, Canada. T2P 5A2 Tel: (403)205-6111 Fax: (403)206-6401 andrew_norrie@veritasdgc.com lustre-discuss-admin@lists.clusterfs.com wrote on 07/05/2004 11:14:39 AM:> On Wed, 2004-04-28 at 18:41, Andrew_Norrie@veritasdgc.com wrote: > > > > Using v1.0.4, pre-patched kernel, RH8.0. > > I''ve successfully configured the 0-config setup with 1 filesystem, but > > cannot get a file system to mount when 2 or more file systems are > > configured. > > Are 2 or more file systems possible with 0-config or is there something > > wrong with the following configuration? > > What you have looks correct, but I would suggest trying something > slightly different, because I don''t think the multi-FS-plus-zeroconf > case has been exceptionally well-tested. > > > [root@node0 lustre]# cat config.sh > > --add node --node node0 > > --add net --node node0 --nid node0 --nettype tcp > > --add mds --node node0 --mds mds1 --fstype ext3 --dev /tmp/mds1 --size > > 100000 > > --add lov --lov lov1 --mds mds1 --stripe_sz 1048576 --stripe_cnt 1 > > --stripe_pattern 0 > > --add node --node node10 > > --add net --node node10 --nid node10 --nettype tcp > > --add ost --node node10 --lov lov1 --fstype ext3 --dev /scratch/ost1--size> > 100000 > > --add node --node client > > --add net --node client --nid client --nettype tcp > > --add mtpt --node client --mds mds1 --lov lov1 --path /mnt/lustre1 > > --add mds --node node0 --mds mds2 --fstype ext3 --dev /tmp/mds2 --size > > 100000 > > --add lov --lov lov2 --mds mds2 --stripe_sz 1048576 --stripe_cnt 1 > > --stripe_pattern 0 > > --add ost --node node10 --lov lov2 --fstype ext3 --dev /scratch/ost2--size> > 100000 > > --add mtpt --node client --mds mds2 --lov lov2 --path /mnt/lustre2 > > What if you remove this last line, and add: > > --add node --node client2 > --add net --node client2 --nid ''*'' --nettype tcp > --add mtpt --node client2 --mds mds2 --lov lov2 --path /mnt/lustre2 > > and then mount fs1 with: node0:/mds1/client > and fs2 with: node0:/mds2/client2 > > Does that help? > > -Phil > > > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss@lists.clusterfs.com > https://lists.clusterfs.com/mailman/listinfo/lustre-discuss
Andrew_Norrie@veritasdgc.com
2006-May-19 07:36 UTC
[Lustre-discuss] 0-config with multiple file systems having problems
Hi, Using v1.0.4, pre-patched kernel, RH8.0. I''ve successfully configured the 0-config setup with 1 filesystem, but cannot get a file system to mount when 2 or more file systems are configured. Are 2 or more file systems possible with 0-config or is there something wrong with the following configuration? The setup is: node0 has 2 mds devices, 1 for each file system. node10 has 2 ost''s (1 ost per file system) node2 is the client The error when attempting to mount the first file system is: [root@node2 lustre]# mount -t lustre -o nettype=tcp node0:/mds1/client /mnt/lustre1 /sbin/mount.lustre: Invalid argument and the errors in /var/log/messages on node2: Apr 28 17:02:40 node2 kernel: LustreError: 1115:(obd_config.c:101:class_attach()) obd OSC_node0_OST_node10_2_MNT_client_2-ea5af800 already attached Apr 28 17:02:40 node2 kernel: LustreError: 1115:(llite_lib.c:455:lustre_process_log()) class_config_parse_llog failed: rc = -17 Apr 28 17:02:40 node2 kernel: LustreError: 1115:(llite_lib.c:540:lustre_fill_super()) Unable to process log: client Apr 28 17:02:40 node2 kernel: LustreError: 1115:(obd_config.c:511:class_process_config()) no device for: MDC_node0_mds1_MNT_client-ea5af800 Apr 28 17:02:40 node2 kernel: LustreError: 1115:(llite_lib.c:455:lustre_process_log()) class_config_parse_llog failed: rc = -22 Apr 28 17:02:40 node2 kernel: LustreError: 1115:(llite_lib.c:603:lustre_fill_super()) Unable to process log: client-clean ''lwizard'' was used to create config.xml and config.sh. lconf on node10 (ost server) and node0 (mds server) runs without errors. config.sh created from lwizard looks like: [root@node0 lustre]# cat config.sh --add node --node node0 --add net --node node0 --nid node0 --nettype tcp --add mds --node node0 --mds mds1 --fstype ext3 --dev /tmp/mds1 --size 100000 --add lov --lov lov1 --mds mds1 --stripe_sz 1048576 --stripe_cnt 1 --stripe_pattern 0 --add node --node node10 --add net --node node10 --nid node10 --nettype tcp --add ost --node node10 --lov lov1 --fstype ext3 --dev /scratch/ost1 --size 100000 --add node --node client --add net --node client --nid client --nettype tcp --add mtpt --node client --mds mds1 --lov lov1 --path /mnt/lustre1 --add mds --node node0 --mds mds2 --fstype ext3 --dev /tmp/mds2 --size 100000 --add lov --lov lov2 --mds mds2 --stripe_sz 1048576 --stripe_cnt 1 --stripe_pattern 0 --add ost --node node10 --lov lov2 --fstype ext3 --dev /scratch/ost2 --size 100000 --add mtpt --node client --mds mds2 --lov lov2 --path /mnt/lustre2