Fibre Channel (SAN)

Reply
Occasional Contributor
Posts: 6
Registered: ‎02-21-2011

Configuring Multipathing an dredundancy

Hi;

I am thrown into the deep end to configure redundancyon a live environment. Basically, there is one enclosure C7000 with mixture of blades some servers are physical and some are virtual.

They connect through the interconnect in the back of the enclosure to two brocade silkwork 200. Each brocade then connects to the EVA 4400.

Currently, there are two interconnect switches with 4 port. Port 1 of each interconnect is connected to a port in brocade fabric then zoned. Someone has configured something like 15 zones for 7 servers which seems too extreme.

I want some guidance on:

1: Understand how the zoning works with the interconnect

2: How to connect the interconnect to the brocades to ensure redundancy

3: two cables to the EVA each from a brocade swith seems to me not highly redundant. Thinking should have at least two connection to each brocade fabric

4: Design an efficient zoning and make it resilient

The SAN is an EVA 4400 with two Storage Processors. I am really against the time and would appreciate some quick help. I use the Webtool to do all the zoning and management. There is no documentation, is there anything I should look out for ?

I attach a rough sketch of the current environment

Thanks

Valued Contributor
Posts: 931
Registered: ‎12-30-2009

Re: Configuring Multipathing an dredundancy

First off as you are in an a HP shop, look at their san design guide.

Just follow the link http://bizsupport.austin.hp.com/bc/docs/support/SupportManual/c00403562/c00403562.pdf

1 What you describe looks like an Virtual Connect module to me, see link for 8 G model http://h18004.www1.hp.com/products/blades/components/ethernet/20-8gb-f/index.html . Those are managed thru the Onboard Administrator module. Depending on your configuration the only interaction ranges from replaced flogi's with fdsic'd only to publish virtual wwn's (effectively replacing the wwn on the mezzazine card). If you're not sure and don't know how to access management, take a look at the module in question. It should be printed on the outside whether its an Virtual Connect or SAN switch, A bill of materials or delevery bill will do as well.

2 Switch or Virtual Connect, both should be attached to a single SAN switch. What you could do is to add another ISL for redundancy (perhaps trunk it for more bandwith).

3 Your EVA should have two cables per Storage Processor installed, that makes a total of 4 cables. Each SP should be connected to either switch.

4 I do my zoning per single initiator and single target, whihc is a HP best practice if I'm not mistaken. Please reference to the design guide. With 4 servers it will created 4 zones per fabric, 8 in total.

Occasional Contributor
Posts: 6
Registered: ‎02-21-2011

Re: Configuring Multipathing and redundancy

Hi;

Thanks for the links, I have scanned through and got some useful information. My immediate concern is to make sure that I set up redundancy as someone removed a cable from one of the VC and it crashed the lot.

The thing that really bothers me is to understand how the Server WWN / VC WWN / Brocade sw WWN / Storage PWWN all fit together to give redundant or multi pathing when zoning.

The way I see it is :

On Brocade SW1

Zone_Server1(target Controller 1 PWWN, WWN1 for Server1 ) and add the VC1 port1 WWN as member to the zone. This will be path 1 to the target LUN1

On Brocade SW2
Zone_Server1 (Target Controller 1 PWWWN, WWN2 for Server1) and add VC2 port 1 WWN as member to the zone. This will be path 2 to LUN1


But how many zones do I have to create and what is best practice for zoning (as I said I am on deep end here)

As for your point below:

2 Switch or Virtual Connect, both should be attached to a single SAN switch. What you could do is to add another ISL for redundancy (perhaps trunk it for more bandwith).

VC1 port 1 and VC2 Port 1 should both go to the same Brocade switch. I need to cable up VC1/2 Port 2 to Brocade SW2? wouldn't this give you a SPOF on the VC end ?

Would appreciate it if you could illustrate with a differenct colour on the word doc I attached. I attach the current config, I would not have been so cautious if this was not a live set up.

Valued Contributor
Posts: 931
Registered: ‎12-30-2009

Re: Configuring Multipathing and redundancy

The way I see it is :

On Brocade SW1

Zone_Server1(target Controller 1 PWWN, WWN1 for Server1 ) and add the VC1 port1 WWN as member to the zone. This will be path 1 to the target LUN1

On Brocade SW2
Zone_Server1 (Target Controller 1 PWWWN, WWN2 for Server1) and add VC2 port 1 WWN as member to the zone. This will be path 2 to LUN1


But how many zones do I have to create and what is best practice for zoning (as I said I am on deep end here)

Put simply and on the assumption you have an flat fabric with nu traffic isolation.

In zoning the only things that matters is for the end device (initiators and targets ) which have to see each are in the same zone.

Exmp. you have an EVA and a blade server, thus the wwpn's of the EVA and wwn (VC or real) of your blade are in one zone PER fabric. repeat for each blade.

No need to put wwn's of VC module or switches in zones.

As for the question if an VC should be connected tot two switches, the answer is "it depends", but usually a VC is connected to one Fabric (switch) only.

With regards to your zoning config. It looks like you have 6 bladeservers one EVA and one MSA.

It doen't look that weird but could perhaps do with a little housekeeping (upercase versus lowercase that sort of thing, remove unused zones).

If alias information is correct for MSA EVA and Blades

1> fabric A alias comvirt06 is the same as blade01_1

2> fabric A alias blade01_1 has an second (probably the wwNn) configured, only wwPn is necessary.

3> fabric B missing something that looks like comback04_2 in fabric A

4> don't know what comvirt* should be, can you

5> remove unneccesary or unused information, for example zone COMVIRT01_MSA_A is defined but not effective.

Occasional Contributor
Posts: 6
Registered: ‎02-21-2011

Re: Configuring Multipathing and redundancy

Dion

Great, it is becoming clearer with loads of quick reading and advise.

Have found out that to have failover, I need to set up a second port on the CV1 and connect it to a second port of the same Brocade. This is done through VC Management console. Apparently, HP does not support a mesh set up if you have multiple VCs. Some say it also depends on the VC firmware version.

Now to make matter little more confusing for me, Blade 1 is a VMhost and has Virtual Machines ( comvert*) but I don't think this a real issue.

Comback04 is a physical server with two HBAs each connecting to a Fabric at the moment and zoning needs to be the same on both swicthes

I would like to recreate the Zones and name them appropriatly as they do not fit with the physical. I am also not 100% about the members in each zone.

Zoning attempt 2: 

Step1 :

Create Alias for VC1 and add VC1 WWN

Switch 1                                                        Switch2

Alias1 / associate WWN of the VC1 port 1        Alias1 /  associate WWN VC1 port 1 

Zone1 on Switch 1                              Zone 1 on Switch 2

Z1(Traget controller1, Alias1)                Zone1(target controller1, Alias1)

Zone1 members:  Target controller 1 / either VC1 or server1 WWN not sure here ??? becuase the VC port will register the server WWN

In this set up I still can see how to achieve multipathing and redundancy ? If HBA1 on server 1 goes down then I lose the storage ?????

Valued Contributor
Posts: 931
Registered: ‎12-30-2009

Re: Configuring Multipathing and redundancy

Dion

Great, it is becoming clearer with loads of quick reading and advise.

Have  found out that to have failover, I need to set up a second port on the  CV1 and connect it to a second port of the same Brocade. This is done  through VC Management console. Apparently, HP does not support a mesh  set up if you have multiple VCs. Some say it also depends on the VC  firmware version.

Yes it's a bit like another ISL between your VC and Switch.

Now  to make matter little more confusing for me, Blade 1 is a VMhost and  has Virtual Machines ( comvert*) but I don't think this a real issue.

Those VM's happen to have their own wwn? NPIV on NPIV .

Comback04  is a physical server with two HBAs each connecting to a Fabric at the  moment and zoning needs to be the same on both swicthes

Check your zoning as I didn't find anythin that looks like comback04 in on of the fabrics. That doesn't mean its not there, but merely because of me being unfamiliar with your environment

I  would like to recreate the Zones and name them appropriatly as they do  not fit with the physical. I am also not 100% about the members in each  zone.

Zoning attempt 2:

Step1 :

Create Alias for VC1 and add VC1 WWN

Switch 1                                                        Switch2

Alias1 / associate WWN of the VC1 port 1        Alias1 /  associate WWN VC1 port 1

Zone1 on Switch 1                              Zone 1 on Switch 2

Z1(Traget controller1, Alias1)                Zone1(target controller1, Alias1)

Zone1  members:  Target controller 1 / either VC1 or server1 WWN not sure here  ??? becuase the VC port will register the server WWN

Only zone the wwn's that relate to FC-HBA's  (servers/ initiators) and storage.

In  this set up I still can see how to achieve multipathing and redundancy ?  If HBA1 on server 1 goes down then I lose the storage ?????

No that should not be the case.

I've attached an drawing how everything should be connected.

VC1 and Switch 1 are the blue fabric

VC1 and Switch 1 are the red fabric

Every blade (should) have an dual ported fc hba.

Each of those ports should be routed to either VC. This is do by internal wiring and or setup usem the VC and OA management software by defining profiles.

Each fabric is managed seperatly and independent.

Each EVA SP is wired to red and blue fabric.

You should have HP EVA DSM MPIO module installed on windows host as your multipath solution. Other OS like RHEL uses the dev-mapper for this.

If everything is working and setup correctly the only way to lose your storage is if the EVA dies or your HBA.

The loss of a single connection, switch, VC, EVA SP should have no impact on the services you provide.

Occasional Contributor
Posts: 6
Registered: ‎02-21-2011

Re: Configuring Multipathing and redundancy

Dion;

You just came to the point I have been struggling with all along. Your picture shows only one connection between VC and Switchx (which are the brocade in my case). I think I need VC1 connecting to both switches and VC2 connecting to both switches.

Zoning, I would create a Zone for VC1 and a zone for VC2 with the same target. The only thing that need to be figured out is how do the server HBAs map VC ports.

I think if I can figure out these bits, I am there in theory. But the proof is in the pudding.

Just aside: I need two extra cables as the current Switches have a single connection to the target HBA

Valued Contributor
Posts: 931
Registered: ‎12-30-2009

Re: Configuring Multipathing and redundancy

You just  came to the point I have been struggling with all along. Your picture  shows only one connection between VC and Switchx (which are the brocade  in my case). I think I need VC1 connecting to both switches and VC2  connecting to both switches.

By bad, fixed it. And no VC1 remains connected to switch1 only

Zoning, I would create a Zone for VC1 and a zone for VC2 with the same  target. The only thing that need to be figured out is how do the server  HBAs map VC ports.

Explain to me what you mean with VC1. Is it the module itself, is it the wwn it gives to the server. As i read it it means VC module to me, thus a zone with VC1 in it makes no sense as its neither a target not initiator.

Can you make a drawing with what you have and another one with what you want?

Occasional Contributor
Posts: 6
Registered: ‎02-21-2011

Re: Configuring Multipathing and redundancy

Dion;

You just came to the point I have been struggling with all along. Your picture shows only one connection between VC and Switchx (which are the brocade in my case). I think I need VC1 connecting to both switches and VC2 connecting to both switches.

Zoning, I would create a Zone for VC1 and a zone for VC2 with the same target. The only thing that need to be figured out is how do the server HBAs map VC ports.

I think if I can figure out these bits, I am there in theory. But the proof is in the pudding.

Just aside: I need two extra cables as the current Switches have a single connection to the target HBA

Valued Contributor
Posts: 931
Registered: ‎12-30-2009

Re: Configuring Multipathing and redundancy

That's the same reply as you gave in http://community.brocade.com/message/16038#16038

Your mistake???

Join the Community

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

vADC is now Pulse Secure
Download FREE NVMe eBook