Fibre Channel (SAN)

Reply
Occasional Contributor
Posts: 5
Registered: ‎09-16-2016

Cabling & fabric decision

[ Edited ]

Hello everyone,

 

I’m kind of new in the storage world. I’m still learning.

 

Today I have a HP C7000 blade enclosure with 2 x HP B-series 8/12c SAN Switch BladeSystem c-Class. These are connected to our EVA 6350 controllers (see the attached diagram how it's connected today). The EVA has 8 FC ports, 4 on each controller.

 

I want to install 1 more C7000, with same model of brocade SAN switches to same storage. What’s the best practice to do with cabling, fabrics, ISL?

  

Thanks in advance!

Contributor
Posts: 66
Registered: ‎12-24-2015

Re: Cabling & fabric decision

[ Edited ]
Hi!
The best way it's buy additional switch like Brocade 300.
Connect your storage directly to switch, change blade switches mode to access gateway and make your SAN in redundant topology.

Another way you can commutate your storage to each switch.
Chassis1 bay3 - storage port1 ctrl1;
Chassis1 bay4 - storage port1 ctrl2;
Chassis2 bay3 - storage port2 ctrl2;
Chassis2 bay4 - storage port2 ctrl1.

Establish isl trunk port (if you have needed licence) between:
Chassis1 bay3 - Chassis2 bay3;
Chassis1 bay4 - Chassis2 bay4.

In your case you can connect fp3 storage ports to SW3 and fp4 storage ports to SW4.
About ISL connections...
You will need make them if:
1. You have masking on storage for all ports.
2. You would create for example, vmware clusters between chassis.
3. Some servers generate high workload traffic and you want spread the load.
Occasional Contributor
Posts: 5
Registered: ‎09-16-2016

Re: Cabling & fabric decision

[ Edited ]

Thanks for your replied!

 

Additional switch is that for scalability in the future?

I don't have budget for that now, but should consider it.

 

Your second suggestion was my primary way to go, but I was even unsure. I have the switches licensed with ISL (not trunking I guess). If I understand correctly I get 2 fabrics (left & right switches).

 

Your third suggestion. Does it mean I should not use ISL? Because I connect SW3 and SW4 directly to new storage ports. What are the pros and cons for this and second suggestion? 

 

"About ISL connections" was it generally speaking or to the third suggestion. I don't understand about the masking on storage for all ports. I'm going to have VMware cluster between the chassis. I don't have the data for workload right now so can't answer that. 

 

Sorry for my stupid questions.

 

 

Contributor
Posts: 66
Registered: ‎12-24-2015

Re: Cabling & fabric decision

Hi!

"About ISL connections" was to the third suggestion.
My third suggestion mean that you may use or not use ISL.
ISL not always good for enviroment.
It combine best practices for storage and SAN. When you make mapping on your storage and not use ISL in SAN you must map luns for chassis1 strictly on fp1 and fp2 storage ports. And for chassis2 strictly on fp3 and fp4 storage ports.

In use ISL you can map luns on all storage ports. But keep in mind about wwn oversubscription on storage ports and additional hop throught isl between switches.

In this two ways do not forget about the same lunID mapping for blade on storage.
Occasional Contributor
Posts: 5
Registered: ‎09-16-2016

Re: Cabling & fabric decision

[ Edited ]

Hi!

 

Thanks.

 

If I go with the third suggestion without ISL, does it mean I can't share LUNs between the chassis? 

 

I must say I like the second suggestion. I have draw the cabling (see file attached), is it OK? 

 

 

Contributor
Posts: 66
Registered: ‎12-24-2015

Re: Cabling & fabric decision

Kolg!

 

I attach my vision on your enviroment.

 

Otherwise, it isn't metter have you isl between switches or not.

It's main only for your vision and design of infrastructure.

Occasional Contributor
Posts: 5
Registered: ‎09-16-2016

Re: Cabling & fabric decision

[ Edited ]

Wow, I really appreciate you taking your time

 

Thanks!

 

I think I understand what you mean with LUN masking. In the EVA we have created each host as a alias, and added HBA WWN (2) of the server and present the LUNs individual. We have done that with all servers. 

 

Depending which way I choose, does it matter how we LUN masking? Your first way (left one) is how we do it today. The second way seems like I create a alias with all HBA WWN for Bay1 (each chassis) or all blade servers? 

 

 

Contributor
Posts: 66
Registered: ‎12-24-2015

Re: Cabling & fabric decision

[ Edited ]

Not at all Smiley Happy

 

It does not metter which way you choose. It's only sight on architecture of your solution.

 

In additional what would I like says:

You must separate zonning on SAN and LUN masking on storage.

Zonning just separate communications between server's HBA and storage controller port.

From my experience I strongly recommended in your small infrastructure make "single initiator - single target" zonning. Where each alias contain only one wwpn. This is more difficult for administer, but it is more easy in future for troubleshooting and elastic configuration.

 

In storage inviroment by LUN masking you separate access for logical essences, such vdisk or LUN or something else.

You can organize access for each server (for two WWPNs, how you do it today) or for server's set (some WWPNs of servers which includes in common cluster). I think what storage's vendor have his personal view for best practice masking.

But in general you must remember, if you create for each server it's own masking, servers in common cluster must have the same LunID for mapping vdisks!

If you create on your storage system alias which contain all server's wwpns of common cluster, you don't need to remember about LunID. It will be the same by default.

Contributor
Posts: 66
Registered: ‎12-24-2015

Re: Cabling & fabric decision

The second way seems like I create a alias with all HBA WWN for Bay1 (each chassis) or all blade servers?

 

No, it means only HBA WWN of bays in VMWare Cluster

Occasional Contributor
Posts: 5
Registered: ‎09-16-2016

Re: Cabling & fabric decision

Learned a lot. Big thanks.

 

About zonning, and single initiator - single target. I maybe missunderstand your picture but I see it like this:

 

With the ISL way

 

Alias: CHASSIS1_BAY1_HBA1 = WWPN1

Alias: CHASSIS1_BAY1_HBA2 = WWPN2

 

Zone:

CHASSIS1_BAY1_HBA1

CONTROLLER FP 1 3 (2 WWN in same alias)

 

Zone:

CHASSIS1_BAY1_HBA2

CONTROLLER FP 2 4 (2 WWN in same alias)

 

Alias: CHASSIS2_BAY1_HBA1 = WWPN1

Alias: CHASSIS2_BAY1_HBA2 = WWPN2

 

Zone:

CHASSIS2_BAY1_HBA1

CONTROLLER FP 2 4 (2 WWN in same alias)

 

Zone:

CHASSIS2_BAY1_HBA2

CONTROLLER FP 1 3 (2 WWN in same alias)

 

Is this correct? 

 

What will happend if I create for each server it's own masking and the LunID isn't the same? 

 

 

Join the Community

Get quick and easy access to valuable resource designed to help you manage your Brocade Network.