Virtual array 7110 and SAN

[jpcast_real]

Hello,

I am trying to connect an VA7110 to an StorageWorks SAN switch to make a fabric... I am having some problems . First of all , when I connect via the serial cable to the array, the installation guide commands do not work correctly.... The manual says "mgr -S 2 -c 1 " however the system says that it does know which is the controller ID or port ID . How can I find the port and controller ID ? Apart from that I suppose that I must connect the fibre cables from the swith to the "HOST" ports of the VA . I wonder why if I connect the cables to the "JBOD" port I can see the 10 disks directly from the ioscan in the HP-UX. Can anyone tell me what should I do in the switch to configure the fabric with the VA. Apart from that I must include a second switch in cascade ... Can anyone give me any recommendations....

[Uwe Zessin]

I have only played with a VA7410 some time ago...

I have used the following command to switch a host-port to 2 GBit/sec:
No Map > mgr -S 2 -c 1 -hp 1
Port Data Rate will be changed to: 2 GB/sec

Change will take effect after a reset.
Do you wish to reset now (yes/no)?:y
Partial Reset.....


HP Virtual Array 7410
Firmware Image 2 selected
Activating DDR DRAM: Success
Copying firmware image to DRAM
yyyyyyyyyxxxxxxxxxx: Initializing:
02 04 06 08 0A 0C 0E 12 16 18 1A 1C 1D 1E 20 22 24 26
No Map >

Did you know that you can request online help through:
No Map > mgr -?


Performs subsystem management operations.

Usage:
mgr -s {shut | start}
mgr -R {full | partial}
mgr -B {autoraid | raid1}
...


You can display the component names with 'dsp -f'. Again, use 'dsp -?' to get online help.

As you can see from the example above, the array I was playing with was not initialized. If I recall correctly that is done with 'vfpfmt'. After that the prompt changes to 'Ready' and you can create any LUNs for your servers.

The servers are to be connected to the 'HOST' ports, correct. The ports marked 'JBOD' are used to link additional disk drive enclosures to your storage array. Do NOT connect any server to it!

What switches do you have? Can you make a litte ASCII chart how you like your fabric and paste it as an attachment?

[Uwe Zessin]

Oops. I see that you have also posted this question to forum 'disk array' which is much better suited - let us continue the discussion there.

[jpcast_real]

I will have two CPDs , 3 kilometers far one from the other one . I will have a four node Service guard clusters with 2 nodes in each CPD . There are 2 Storageworks 2/8 switches in each CPD. I wanted to create two different fabrics . Each fabric will have one switch from each CPD

[Uwe Zessin]

Your switches need unique domain IDs in each fabric - otherwise they will segment.

> Error FABRIC-SEGMENTED, 3, port 15, domain IDs overlap

To change:
fcswa2:admin> switchdisable
fcswa2:admin> configure

Configure...

Fabric parameters (yes, y, no, n): [no] y

Domain: (1..239) [1] 2 <==
BB credit: (1..16) [16]
R_A_TOV: (4000..120000) [10000]
E_D_TOV: (1000..5000) [2000]
Data field size: (256..2112) [2112]
Non-SCSI Tachyon Mode: (0..1) [0]
Disable Device Probing: (0..1) [0]
Suppress Class F Traffic: (0..1) [0]
VC Encoded Address Mode: (0..1) [0]
Per-frame Route Priority: (0..1) [0]

Virtual Channel parameters (yes, y, no, n): [no]
Arbitrated Loop parameters (yes, y, no, n): [no]
System services (yes, y, no, n): [no]
Committing configuration...done.
fcswa2:admin>
fcswa2:admin> switchenable

This is from an older firmware revision, just change the entry marked '<==', however, do not use '8' as this is reserved internally by HP-UX. I assume that you know you need 'long-wave' transceivers in your switches and special fibre cables between your locations?

What das 'CPD' mean, by the way?

[jpcast_real]

Sorry Uwe , CPD means datacenter in spanish...


What does it mean when you say "segment"...

To get a better redundacy , do you think that I should build 2 independt fabrics or unit the 4 switches building ISLs trunking...

[Uwe Zessin]

'segment' in this case means that the two switches do not form a common fabric - no data is flowing between them.

I suggest you connect only a pair of two switches and not build a full mesh. The reason is that with 2 fabrics you have 2 independed namespaces for zoning. An error in one fabric allows the hosts to fail over to the other fabric. That is one of the requirements by HP for a 'high availability' configuration.

Provided that you have enough ports you can also run two links between a pair of switches, but route each link over a different cable tray to your opposite data center.

[jpcast_real]

Thank you very much for everything...

Regards