convert from transparent to multiple-bus failover mode

[Ted Ma_1]

I have a database system consists of a dual hosted DS10 running OS 4.0F and RA8000 FC with dual controller HSG80. I'm currently running transparent failover mode with one SAN switch and want to convert to multiple-bus failover mode so that I can have dure redundant SAN switches.

I already have my databases configured in RA8000 FC. If I convert transparent mode to multi-bus mode, I need first to upgrade my OS to 5.1B. I wonder after doing that, can I still be able to access my databases since the pathes and logical devices may change during the course of upgrading and conversion.

Any information will be highly appreciated.

Ted

I will first upgrade my OS from 4.0F to 5.1B.

since 4.0F OS doesn't support multiple-bus failover mode. The single SAN switch creates a single point of failure. I need to add the second SAN

[Uwe Zessin]

Ted,
there are some things that need to be updated.

In TF-mode units D0..D99 are exclusively presented through controller Port_1 and D100..D199 through Port_2. If you do a 'SHOW CONNECTIONS' you will see that there are connections with UNIT_OFFSET=100 (or values greater than 100).

In MF-mode all units are presented through all controller ports. I guess you will cable the following way:
- 2 fibre channel adapters (FCA) per server
- each FCA goes to a different fibre channel switch
- fibre channel switches are not connected
- both controller Port_1 go to switch-1
- both controller Port_2 go to switch-2

You will then get two more connections. Make sure that they have the proper OPERATRING_SYSTEM set! I suggest you set all connections to UNIT_OFFSET=0.

Tru64 Unix uses the 128-bit 'LUN ID' to map HSG units to its disk names.
> D101 M001
> LUN ID: 6000-1FE1-0009-47C0-0009-0290-1390-0058
> IDENTIFIER = 101

That allows you to move containers around in the unit space like this:
DELETE D101
ADD UNIT D11 M001 DISABLE_ACCESS_PATH=ALL
SET D11 ENABLE_ACCESS_PATH=(S1_A_A1,S1_A_B1,S1_B_A2,S1_B_B2)

However... based on my colleagues experiences I rather suggest you make a new installation of Tru64 Unix and not waste your time with upgrades - especially if this is a cluster!!

[Johan Brusche]

When going to multibus failover in a TruCluster environment, the HSG80 also needs to be set in SCSI-3-mode (now in SCSI-2).
In SCSI-3 mode,logical Unit#0 (ie D0) is reserved for remote console connections, so you better evacuate D0 if it's in use now.

To assist you in correlating the old V4.0F devices names RZxx with the new V5.1B DSKxx names, you can set an otherwise unused field in the disklabel to the RZ-name. I'm referring here to "label:" as shown below.

# disklabel dsk1
# /dev/rdisk/dsk1c:
type: SCSI
disk: RZ1CB-CS
label:
flags: dynamic_geometry
bytes/sector: 512
sectors/track: 113
tracks/cylinder: 20

There is a script called 'clu_migrate' in the V5.x TruCluster subset that does that for you. (Install subset on V4.0F with setld).

Rgds,
Johan.

[Uwe Zessin]

Johan,
where is the information about requiring SCSI-3 from? As far as I can tell, the SAN design guide AA-RMPNL-TE still allows SCSI-2 according to table 30 on page 122. (I know that the general recommendation is to migrate to SCSI-3).

[Johan Brusche]

It's not an absolute need, but a strong recommendation to have a fixed CCL for ease of management, eg use of hsxterm5 -f /dev/cport/scp0. See paragraph 7.9.1.1 in TruCluster Hardware configuration manual,online at:
http://h30097.www3.hp.com/docs/base_doc/DOCUMENTATION/V51B_HTML/ARHGWETE/CHFBRCHN.HTM#sec-setting-up-hsg80

JB.

[Uwe Zessin]

Is 'hsxterm' still supported? I know that 'HSZTERM' is not supported on OpenVMS.

On OpenVMS (which has always required SCSI-3, btw) we could also connect 'through' a disk device. I wonder if that is possible on Tru64 Unix, too...

It is too bad that the command scripter is an extra-cost option :-(