OPS Raw Partition Question

I am trying to install OPS on a Sun Cluster 2.2 setup and have finally got the cluster running and set up my shared veritas disk groups. Those imported as shared to each node nicely and I thought I was close to being done...

My problem is the docs for setting up raw partitions on the shared volumes are rather skimpy (well they're great on concepts, pictures, etc.). I know that I need to have an ascii file with the layout and the locations specified and that I then run the Oracle installer with an environment variable pointing to my file. I also know I should choose an equivalent size to my partitions for future expansion.

So, I currently have this for my disk array:

ops1# vxdisk list

DEVICE       TYPE      DISK         GROUP        STATUS
c0t0d0s2     sliced    -            -            error
c1t0d0s2     sliced    disk01       rootdg       online
c1t1d0s2     sliced    oracled01    oracledg     online
c1t8d0s2     sliced    oracled02    oracledg     online
c1t9d0s2     sliced    -            -            online

where oracledg is the shared array and the two disks are each 18G drives. Because encapsulation failed I had to put the unshared rootdg on the first and fourth disks of the array, as well.

My ascii file currently looks like this (path to be determined):

system1          device/path/op_system1.dbf
users1           device/path/op_user1.dbf
temp1            device/path/op_temp1.dbf
rbs1             device/path/op_rbs1.dbf
indx1            device/path/op_indx1.dbf
tools1           device/path/op_tools1.dbf	(not needed??)
drsys1           device/path/op_drsys1.dbf	(not needed??)
control1         device/path/op_control1.clt
control2         device/path/op_control2.clt
redo1_1          device/path/op_redo1_1.log
redo1_2          device/path/op_redo1_2.log
redo2_1          device/path/op_redo2_1.log
redo2_2          device/path/op_redo2_2.log

and rbs1 needs to be 520M (Multi-purpose db), I've chosen a 600M partition size. So my plan is to create 60 (ea) 600M disks in oracledg (30 on each real disk) and make the assignments in the ascii file.

The questions that I have are the following:

(1) I created the oracledg with vxdiskadmin and see that
they are marked sliced, plus 2 disks have already been created. Do I need to remove the sliced disks (oracled01, oracled02) and recreate them with:

  vxassist -Ufsgen vol op_system1.dbf 600m -logtype=none ??

(2) I put oracledg in /etc/default/vxassist and they say you can
put other stuff in there, but give zero examples. Does someone know the format ??

(3) If I am creating a multi-purpose DB, are the two partions
up there(tools1, drsys1) needed ??

(4) While attempting rootdisk encapsulation, I got a failure
because there were not enough free slices. The guidelines call for (2) free ones. Both c0t0d0s3 and s7 were free, albeit with 500M assigned to them, but not mounted. I also included a small 10M completely unassigned disk space and a very large swap. Does someone know why encapsulation might have failed in this scenario ??

My setup is as follows:

        2 (ea) Sun Ultra 2 enterprise and 1 (ea) D1000 (4 X 18G)

        Sun Cluster 2.2 07/00 VxFS 3.3.3 VxVM 3.0.4

Hopefully, someone can answer some of these. Thank you very much in advance.

-Dirk Received on Fri Jan 05 2001 - 20:57:25 CST