OK, so much for the theory, but how do you configure a DS8K in practice?
You can either use GUI screens, or commands from a command line interface. I'll just discuss CLI commands, as the GUI is supposed to be intuitive. The commands are more powerful and you can pre-define them into a file and run them as a script. If you chose to use the GUI, you need to take the same steps as below, using a suitable GUI screen.
CLI Profiles
When using CLI commands, you should set up a profile file for every DS8K, and specify the DS8K machine data in that profile.
The default profile is in c:\Program Files\IBM\dscli\profile
Among other things, it should contain the IP address of the Hardware Maintenance Console and the DS8K serial number like this
Hmc1: 1.2.12.13
Devid: IBM.2107-75BS072
You can invoke different profiles by using the -cfg parameter
Without a profile you need to specify the hardware device in every command like this
SHOW commands give detailed information about 1 component
LIST commands list out objects for a given type (ranks arrays etc)
MAKE is used to create an object
CHANGE is used to alter an object
REMOVE is used to delete an object
COPY SERVICES commands are used for mirroring and flashcopy and are not discussed here.
Defining 4 new RAID ranks
I'm assuming that your DS8K is installed and licensed, all physical cabling is complete, and your engineer has inserted the disk groups.
To configure a DS8K with two arrays of FC disk and 2 arrays of FATA disk, you go through the following steps (This install has a definite mainframe bias, but where different actions are needed for Open Systems, they are mentioned).
First, define the I/O ports that will be used to communicate with the DS8K. The Setioport command has two parameters, one for the type of communication, which can be one of SCSI-FCP, FC-AL or FICON, and the port number. The number of ports you define should be the same as the number of channels that were cabled up.
Next you create the RAID arrays. To do this, you need to know which Array Sites are free and available for formatting, so you run lsarraysite -l first and note the free Array Sites.
The mkarray command needs two parameters, one to describe what type of RAID you want, and one to specify which Array Site you want to use. RAID types can be 5,6 or 10.
It is always good practice to look at what you just created and make sure the result was what you expected.
lsarraysite -l
lsarray -l
Array State Data RAIDtype arsite Rank DA Pair DDMcap (10^9B)
=====================================================================
A0 Unassigned Normal 5 (6+P+S) S1 - 0 146.0
A1 Unassigned Normal 5 (6+P+S) S2 - 0 146.0
A2 Unassigned Normal 5 (6+P+S) S1 - 0 500.0
A3 Unassigned Normal 5 (6+P+S) S2 - 0 500.0
Note that the DDMcap capacity figure is quoted in units of 10^9 Bytes. This is an IBM Gigabyte, or 1,000,000,000. The more usual figure, and the figure quoted for all Open Systems storage is 2^30 or 1,073,741,824 Bytes. This can cause confusion when calculating capacities.
Next you create the Ranks with the mkrank command and format the array up. In this example the ranks are formatted as ckd, use -stgtype fb for Open Systems
Now you want to create four extent pools, one each of FATA and FC for each server. Each Extent Pool is associated with either Rank Group 0 or Rank Group 1, which in turn are associated with Server0 and Server1. All 4 Extent Pools are CKD, and all 4 pools are given names that show they are CKD, which server they are associated with, and which tier disks they contain.
Now I should have created the extent pools before I created the Ranks, but I can now associated the Ranks with the extent pools with the chrank command.
I want to use some of the Tier3 disks for Space Efficient Pools for backups. repcap is the amount of physical space to reserve for space efficient volumes, while vircap is the amount of virtual space that can be defined as disks.
As these are CKD pools, I need to create Logical Control Units. Open Systems users can skip this step. This command will create 16 LCUs, numbered from B100-B10F.
As each LCU can have 256 volumes, that allows me to address 4096 volumes.
(For information, the B1 is the storage unit identifier and the 00-0F is the logical subsystem identifier. The four combined hex numbers are the LCU address. The Volume addresses will be A000-A0FF, A100-A1FF, etc. . You will also need an IOGEN to define the 16 LCUs and the 4096 volumes to z/OS. The link between LCU numbers and volume addresses is made in the IO gen)
mklcu -qty 16 -id 00 -ss B100
lslcu
An Extent Pool that is associated with Server1 must also be associated with odd numbered LCUs.
Now finally I get to create some volumes. The mkckdvol command needs to know which extent pool to go to for its extents, how big each volume is in cylinders, and how many of each type of volume to add
The commands above show 1 3390-1 being defined (the smallest size available, for use as a GDPS utility volume), 159 mod9s, 48 mod 27s and 48 PAV aliases.
Each CKD volume gets a nickname assigned. The names chosen show the volume size and the assigned extent pool - you can make your own up of course. The #h on the end of the name means to use the hexadecimal volume number as part of the volume name. The final numbers represent volume ranges. 1100 is one single volume. 1101-119F is a range of volumes.
Open Systems specific commands
If you are creating FB volumes, then you need two extra steps;
Create Volume Groups and
Create Host Connections
Server hosts use two different ways to discover disks; SCSI MAP256 or SCSI MASK. You need to know which type your server uses, and you can find this out with the commands
These commands will list all the servers that use each discovery type. If your server is WIN2K, then it uses SCSI MAP256, so the command to create a volume group called VG_W01 is
mkvolgrp -type scsimap256 VG_W01
Next you would create the volumes for that volume group. This command will create 8 50GB volumes in the VG_W01 volume group. Note that the command is mkfbvol.
Finally you need to create Host Connections using the command mkhostconnect. This effectively creates LUN masking, so each server can only see the volumes in its group. Note that a volume can belong to more than one volume group.