recommended way of cabling for DHCI fault tolerance and performance

[jchen522]

I am trying to figure what is the best way for cabling for DHCI setup. 

According to the diagram in "HPE Nimble Storage dHCI and VMware vSphere 6.7u Deployment Guide", page 6, HPE ProLiant DL3x0, the "VM network management + vMotion" ports are assigned to one card, and iSCSI1 and iSCSI2 are connected to another card.  The cables are cross connected to each of the switches.

However, in "HPE Nimble Storage DHCI Solution Network Considerations Guide", beginning on page 9, the HPE ProLiant DL compute nodes, the MGMT and iSCSI1 appears to be assigned to one card, and MGMT and iSCSI2 appears to be assigned to another card.  The cables are not cross connected and connect directly to each its perspective switch.  

What is the supported and correct way of cabling?

Current setup is two sites, two servers in each site. Each site has two M-series switches.  Each site has NS HF40.

[MagdalenaS1]

I don't know how to solve this problem, but I had a similar one and using Firmao helped me

[mamatadesaiNim]

It depends on your NimbleOS version.  

During HPE Storage dHCI deployment on an array running release 6.0.0.0 or later, the deployment tool uses ports 1 and 3 for Management. It uses ports 2 and 4 for iSCSI 1 and iSCSI 2.  This is for increased NIC resiliency.

[jchen522]

Thank you for that info!

Do you know when configuring MLAGs, other than the interconnet between the switches, are there any other MLAGs that I should create toward the Hypervisor hosts or Nimble Storage?

The deployment tool doesn't configure/automate anything on the network switches right?  all MLAG and VLANs port assignments needs to be set before hand correct?

[mamatadesaiNim]

Please refer to this doc https://infosight.hpe.com/InfoSight/media/cms/active/public/HPE_Nimble_Storage_dHCI_and_VMware_vSphere_Deployment_Guide_-_Greenfield_Alletra_Deployment.pdf
As far as MLAGs toward the hosts or array, no, you do not, and should not create any.
The deployment tool CAN configure/automate the network switches If using Aruba 8325 or 8360 switches, but it doesn't do "all" of it.

[jchen522]

Thank you!

I have two sites .  Each site have its own dHCI setup with two hypervisors, two M-series switches and nimble storage array.  When connecting the two sites together, how should the cabling be connected between all 4 switches?

[jchen522]

According to the HPE Nimble Storage DHCI solution network considerations guide, page 12, Design 5 - Configuration with peer persistence, it mentioned each switch at site A must have ISLs to it corresponding switch at site B.  What is configuration are these ISLs exactly between the two sites? MLAG? I thought you can only have two switches in a MLAG configuration. Can these ISLs carry other regular data traffics?

[jchen522]

I have 4 single mode fibre connections capable between the two sites.  How should they be connected? with what protocol?  Thank you!

[RCostin]

@mamatadesaiNim 

Hi

I have the same situation (two sites, each with 1 x Alletra 5030 + 2 x switches + 3 x servers) and I can not find the document specified here

https://infosight.hpe.com/InfoSight/media/cms/active/public/HPE_Nimble_Storage_dHCI_and_VMware_vSphere_Deployment_Guide_-_Greenfield_Alletra_Deployment.pdf

Is it available at some different link ?

Thank you

Costin

[MikeGeroche]

@RCostin 

The current deployment guide for New Installations can be found here:

https://infosight.hpe.com/InfoSight/media/cms/active/public/pubs_HPE_Nimble_Storage_dHCI_Deployment_Guide_for_New_Installations_6_0_x.pdf

[MikeGeroche]

@jchen522 

Regarding Peer Persistence between two dHCI sites:  Right, there must be ISLs between the two sites per the Network Considerations Guide.  Those ISLs will carry data and management traffic between the two dHCI Clusters, and must be in the same layer 2 subnet, as outlined in the HPE Nimble Storage dHCI Solution Network Considerations Guide. 

As stated in the Inter-switch links section on page 5 of that document,  HPE recommends "you include at least two ISLs per switch pair at the highest possible speed to create redundancy and minimize latency".  Hence your single-mode connections meet this recommendation.

Best Practice would be that these ISL interfaces be aggregated together in a multi-chassis link aggregation group (MLAG) running link aggregation control protocol (LACP) in a port-channel.   Note that this MLAG port-channel between the two sites is not the same as the MLAG-IPL between the two switches in each dHCI cluster.  The MLAG-IPL essentially makes the switch pair at each dHCI cluster look like one logical switch, whereas the MLAG port-channel bonds a group of ISLs together and runs LACP, combining multiple physical interfaces into a single logical connection, a port channel.