cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
1434
Views
0
Helpful
44
Replies

First crack at FC using MDS9100

Steven Williams
Level 4
Level 4

I have entered the world of storage networking this past week and need to get MDS9100's up and running for a Lab EMC VNX. I am very versatile in LAN/WAN and also Nexus, just not storage tasks within the Nexus platform. The MDS seems to be NX-OSish and I am familiar with most commands on it. What I have learned so far watching some videos is:

 

Zoning is like ACLs in the R&S world. Zones allow certain ESXi hosts or servers to see certain storage. So I have about 15 luns all part of the same storage pool. So will I have 15 WWPNs? One for each lun?

vSANs are like vlans, I just dont understand when to have multiple vSANs within an environment?

What are some basics I need to know to get storage and servers talking? WWPNs from the servers and storage, I need to know Port type, I assume all port types will be N ports. 

Do MDSs work in like a HA pair? So I connect each MDS to each other and trunk them? I understand there isnt any load balancing in FC like there is in IP when using something like vPC?

 

 

44 Replies 44

jihicks
Cisco Employee
Cisco Employee

Hi Steven,

 

Each interface on storage will have one pwwn.

Each host will have one pwwn.

 

Best practices is to zone one initiator ( host ) to one target ( disk pwwn for instance )

zone name host1_to_disk1 vsan 1

member pwwn xx.xx.xx.xx.xx.xx.xx.xx ( host )

member pwwn  xx.xx.xx.xx.xx.xx.xx.xx (disk )

 

zoneset name production vsan 1

member host1_to_disk1

 

zoneset activate name production vsan 1

 

The best practice for topology is a Fabric A and a Fabric B

Each host connects an interface to each fabric.

Storage is connected to both fabrics.

The fabrics use different VSANs, for example,  say Fabric A is VSAN 10 and Fabric B is VSAN 20.

The fabrics are NOT connected together.

The hosts have multi-pathing software to make use of the two fabrics.

 

VSANs are similar to VLANs.

For one device to see another, they both must be:

   In the same fabric.

   In the same VSAN ( unless you use Inter-VSAN routing which I don't see you using this )

   Zoned together and the zone in the active zoneset ( ACL )

 

On the switch, all ports discussed here are F ( Fabric )  ports.

The host and storage represent N ( Node ) ports.

N ports connect to F ports.

The ports are already auto or Fx on the switch.

Probably will work fine if you connect to them and "no shut" the interface.

 

Hope this helps,

Jim

 

So essentially the EMC storage controller will have two interfaces? or more. Lets say two for simplicity. One will connect to FabricA and the other connects to FabricB. Each port will have their own WWPN with I will code in each fabric and assign an alias like EMC_INT0 and EMC_INT1.

But I have multiple ESXi hosts so the 1 to 1 ratio kind of falls out there no? I would have multiple initiators and one target? 

Hi Steven,

 

Your create a zone for each host.  For this example, we will call the two ESXi hosts, host1 and host2.

Disk1 = pwwn 50:01:73:80:12:c1:02:03
Host1 = pwwn 10:00:00:00:c9:a1:11:22
Host2 = pwwn 10:00:00:00:c9:a1:11:33

zone name host1_to_disk1 vsan 10
member pwwn 10:00:00:00:c9:a1:11:22
member pwwn  50:01:73:80:12:c1:02:03

zone name host2_to_disk1 vsan 10
member pwwn 10:00:00:00:c9:a1:11:33
member pwwn  50:01:73:80:12:c1:02:03

zoneset name production vsan 10
member host1_to_disk1
member host2_to_disk1

zoneset activate name production vsan 1

 

 

 

The other fabric would have a similar configuration with the PWWNs from that fabric and VSAN 20.

 

Best regards,

Jim

 

 

Steven,

 

a couple of specifics as far as VNX is concerned. You always want to have each HBA zoned to ports on SPA and on SPB. For example:

 

Fabric A

esxserver1-hba1-VNX-SPA0

esxserver1-hba1-VNX-SPB2

 

Fabric B

 

esxserver1-hba2-VNX-SPB0

esxserver1-hba2-VNX-SPA2

 

 

@dynamoxxx

Ok I have created all the fcalias for the esxi hosts using the fcalias name command. Now How do I go about getting the pwwn IDs from the VNX storage? I have access to the vsphere client but unsure where to look. 

Once I get that then I can zone them accordingly. Now what about fc ports? How can I make sure that a certain host connects to a certain port? Kind of like a mac sticky in LAN switching.

 

Do you have access to Unisphere ?  You can look in Unisphere to see port WWNs so then when you look on the switch and run "show flogi database" you can validate which VNX ports are logged-in where.  At that point you will be ready for zoning.

@dynamoxxx

No access to Unisphere. Also nothing plugged into the MDS's at this point until the outage window comes so I like have to have this thing ready to go when we flip over the hosts and the storage. We are separating lab from prod.

so how do you expect to create fsalises for VNX and then create zones if you dont' know what WWN to use for VNX ports ?

@dynamoxxx

Ok I have access to unisphere...but I only know data ONTAP so where do I find these?

in Unisphere go to Settings > Network.  There you will see all ports on SPA and ports on SPB.  You need to decided which ports will connect to which fabric and then you can do your pre-zoning ahead of time. The WWN are lised in IQN/WWN column

@dynamoxxx

I assume you would do SPA to MDS 1 and SPB to MDS 2 seems like the common way I would assume. 

Looks like I have 8 ports per SP?

A-0 and B-0 say Mirror view? What does that mean?

How do I know if all of these are in use and need to be aliased and zoned?

actually you want SPA and SPB to each MDS. Look at the at example i provided earlier.

 

Looks like your model was shipped with 8 ports per SP, some have 4 per SP and some have up to 20.

MirrorView ports are used for VNX to VNX replication but they can also be used for regular host connectivity.

 

How to determine which ones in use ? In Unsiphere click on Hosts > Initiators. There you will see everything that has been connected to this VNX and what ports are being used.

 

@dynamoxxx

Wow so this could be a long line of configuration when you start to zone on a 1:1 basis on each MDS. 

Also why do these pwwn values seem way longer than my ESXi host values?

 

you could enable SmartZoning on your VSANs and then place multiple targets in the same zone. 

 

WWN is long because it consists of WWNN:WWPN  (node name and port name). You will always zone to WWPN (last part of that long string)

 

 

@dynamoxxx