recipe for gfs whit parallel scsi and gulm

--0-1061452068-1031703380=:22726
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
Content-Disposition: inline

Hi,

I'm begginer with GFS and i need some helps please.
I've gfs 5.1.0 to evaluate which software license
number is BE-0005-10001015.
I want to set up GFS on cluster of nodes with parallel
SCSI with IP lock server using gulm lock server but
neither of your recipes match with that. Then i'm
attempt create one specific recipe groundoing in yours
recipes in section 5.3(GFS with Fibre Channel and gulm
lock server) and 5.5(GFS with parallel scsi and
LOCK_SWDMEPD) to do why i want, but i have some doubt:
1.Why the pool module is not loaded in recipe 5.5?
2.Using IP-based DMEP, all pooldevices should have a
weight of zero. The same role apply to IP-based GULM?
3.Why the i/o fence daemon are not started in the
nodes in the recipe of section 5.3?
   
Please, Could you see the file attached with the
description of the recipe and localized questions i'm
trying to create, to give me the feedback if it's
correct.

Other question:
Who is the correct way to stop the gfs before do
reboot or shutdown the nodes?

Thanks for your helps.
Manuela Fortes.

_______________________________________________________________________
Yahoo! PageBuilder
O super editor para criação de sites: é grátis, fácil e rápido.
http://br.geocities.yahoo.com/v/pb.html
--0-1061452068-1031703380=:22726
Content-Type: text/plain; name=recipe
Content-Description: recipe
Content-Disposition: inline; filename=recipe

Attempt of to create a recipe for GFS with parallel scsi and gulm lock server.
I have some doubt in the following points: 2, 3, 5 and 11.

Setup Procedure
1. Verify that the two partitionsfrom the RAID device have been found.

2. Load kernel modules on host a, b and c. 
   Question --> (*)why the pool modules is not loaded in recipe 5.5 ?
   host-a$modprobe pool  
   host-a$modprobe gfs  
   host-a$modprobe lock_gulm
  
   host-b$modprobe pool  
   host-b$modprobe gfs    
   host-b$modprobe lock_gulm  

   host-c$modprobe pool  
   host-c$modprobe gfs    
   host-c$modprobe lock_gulm 

3. Create a pool for the file system. on host-a create a file named pool0.cf. 
   Question --> (*) Using IP-based DMEP, all pooldevices should have a weight of zero. The same role apply to IP-based GULM?
   host-a$cat >pool0.cf
   poolname pool0
   subpools 1
   subpool 0 0 1 gfs_data
   pooldevice 0 0 /dev/sdb2 0   
   

4. Create a pool0.
   host-a$pool_tool pool0.cf

5. Create a second pool for the GFS configuration information. On host-a create a file named pool0cidev.cf.
   Question --> (*) The same question of point 3.
   host-a$cat >pool0cidev.cf
   poolname pool0_cidev
   subpools 1
   subpool 0 0 1 gfs_data
   pooldevice 0 0 /dev/sdb1 0 

6. Create pool0_cidev.
   host-a$pool_tool pool0cidev.cf

7. Load the two new pools on all nodes.
   host-a$pool_assemble
   host-b$pool_assemble
   host-c$pool_assemble

8. Create the file system.
   host-a$gfs_mkfs -p lock_gulm -t xptoFileSystem -j 3 /dev/pool/pool0

9. Set up the gulm configuration file.
   host-a$cat >gulm.cf
   gulmcluster:
   node = lockserver
   port = 40040
   heartbeat_rate = 30
   allowed_misses = 2

   locktable:
   name = xptoLockTable
   node = lockserver
   filesystem = xptoFileSystem
   port 42040
   
   node:
   name = lockserver
   ipaddr = 172.16.1.1
   fm = manual 172.16.1.1

   node:
   name = host-a 
   ipaddr = 172.16.1.2
   fm = manual 172.16.1.2


   node:
   name = host-b 
   ipaddr = 172.16.1.3
   fm = manual 172.16.1.3

   node:
   name = host-c 
   ipaddr = 172.16.1.4
   fm = manual 172.16.1.4

   fence:
   name = manual
   agent = fence_manual

10. Write the config file to the cidev.
    gfs_conf -c -d /dev/pool/pool0_cidev -f gulm.cf

11. Start the i/o fencing daemon on each node.
    Question --> (*) why this step is not done in the recipe of section 5.3 ?
    host-a$fenced
    host-b$fenced
    host-c$fenced 

12. Start the gulm lock server on the node designated for this purpose.
    lockserver$lock_gulmd /dev/pool/pool0_cidev

13. Install filesystem license.
    host-a$mount -t gfs -o hostdata=host-a:/dev/pool/pool0_cidev /dev/pool/pool0 /gfs
    host-a$gfs_tool setlicense /gfs </path/to/license_file>

14. Mount the the GFS file system on node b and c using /gfs as the mount point.
    host-b$mount -t gfs -o hostdata=host-b:/dev/pool/pool0_cidev /dev/pool/pool0 /gfs
    host-c$mount -t gfs -o hostdata=host-c:/dev/pool/pool0_cidev /dev/pool/pool0 /gfs


--0-1061452068-1031703380=:22726--

_______________________________________________
gfs-users mailing list
[email protected]
http://lists.sistina.com/mailman/listinfo/gfs-users
Read the GFS Howto:  http://www.sistina.com/gfs/Pages/howto.html