MSA Storage
cancel
Showing results for 
Search instead for 
Did you mean: 

Datastore configuration of LUN's located on HP MSA2040 storage ...

 
DiWi
Occasional Contributor

Datastore configuration of LUN's located on HP MSA2040 storage ...

Hello everybody,

 

I am trying to phrase my question as clearly as possible by describing my environment ...

 

We use 2x HP DL380p Gen8 Server with 2x 10GBit/s iSCSI ports on each Server. The storage is an HP MSA2040 with 4x 10GBit/s iSCSI ports on each Controller (Dual Controller installed).

 

The SAN of the HP MSA2040 storage ist connected as follows ...

 

iSCSI Port 1 with IP 192.168.254.210 (first HP DL380p Gen8) is connected to iSCSI Port A1 with IP 192.168.254.110 (Controller A of MSA2040)

iSCSI Port 2 with IP 192.168.254.211 (first HP DL380p Gen8) is connected to iSCSI Port B1 with IP 192.168.254.111 (Controller B of MSA2040)

iSCSI Port 1 with IP 192.168.254.215 (2nd HP DL380p Gen8) is connected to iSCSI Port A2 with IP 192.168.254.115 (Controller A of MSA2040)

iSCSI Port 2 with IP 192.168.254.216 (2nd HP DL380p Gen9) is connected to iSCSI Port B2 with IP 192.168.254.116 (Controller B of MSA2040)

 

I just configured all iSCSI ports of the MSA2040 with corresponding IP addresses, I configured one vSwitch with 2 VMkernel ports (each with 2x 10GBit/s vmnics). In one iSCSI VMkernel port, one vmnic is unused, in the other iSCSI VMkernel port the other adapter is unused in NIC teaming, so that I am able to configure VMkernel Port Bindings under Network Configuration by selecting my 2 VMkernel ports in the VMware iSCSI Software Adapter.

 

After entering the first IP address 192.168.254.210 under Dynamic Discovery, I can see all 8 iSCSI Ports with their corresponding IP Address under Static Discovery. I did not enter any additional IP Address under Dynamic Discovery.

 

With this configuration, I can see 2 host entries or IQN's under Hosts in the MSA2040 overview. Is that sufficient? Do I Need to add more Hosts manually or is the configuration complete?

 

Now, when creating a Vdisk and one LUN and mapping this LUN to all iSCSI ports of the MSA2040 and after a RESCAN of the iSCSI Software Adapter, I can see one additional Device and 2 additional Paths. Selecting Paths shows me the mapped LUN twice, one LUN is 'Active I/O', the other is only 'Active'.

 

Now I select Storage in vSphere and add the LUN as a Datastore. Everything fine until now. When I manage the path of this LUN, I can see the Storage Array Type set to VMW_SATP_ALUA and the selected Path Selection Policy set to MRU (which is the Default). Reading the 'HP MSA2040 Storage Configuration and Best Practices for VMware vSphere', I know that Round Robin is supported, so I decided to select 'Round Robin' as Path Selection Policy, but the Status doesn't Change, one iSCSI target is 'Active (I/O)', the other stays at only 'Active'.

 

Is this correct? I thought both iSCSI target would Switch to 'Active (I/O)' as I have seen in this Video ...

 

https://www.youtube.com/watch?v=Mu-HyD3E3cw

 

... from vmsources.com?

 

Would I need to connect the second iSCSI port of one ESXi host Adapter to another iSCSI port of the same Controller on the MSA2040 to get 2 'Active (I/O)' paths? (I know this is not recommended but just a question).

 

Best regards,

DiWi

 

1 REPLY 1
Steven Clementi
Honored Contributor

Re: Datastore configuration of LUN's located on HP MSA2040 storage ...

The video you are referring to... depicts what an EMC Storage Array might look like to VMware, not an MSA2040.

 

Different Storage Arrays have different operational configurations and you cannot expect to get the exact same results from one array as you get from another.

 

While the MSA 2040 is an Active/Active type array... one controller is still handling the i/O so yes, if you attached both ports of one server to both ports of one controller... you would see both paths as Active (I/O).

 

This also destroys your controller redundancy which is why you would not do it.  Instead... you should get a 10GbE switch (or 2) and configure the networking properly if you want dual active "in use" paths.

 

Additionally, for future datastores that you create... if you want to skip having to set the disk for Round Robin, you can set the default mode via command line...

 

esxcli storage nmp satp set --default-psp=VMW_PSP_RR --satp=VMW_SATP_ALUA 

 

Any array that you add that uses the ALUA SATP will automatically set to Round Robin.

 

 

Steven Clementi
HP Master ASE, Storage and Clustering
MCSE (NT 4.0, W2K, W2K3)
VCP (ESX2, Vi3, vSphere4, vSphere5)
RHCE
NPP3 (Nutanix Platform Professional)