Comware

 View Only
  • 1.  DRNI compatibility with non-HPE devices

    Posted Jul 17, 2025 05:37 AM

    Hi,

    I am new to HPE networking and I am looking into using HPE Networking Comware Switch 5980 or something similar to be used as the TOR switches for a cluster of hyperconverged infrastructure servers (Nutanix) which support LACP.

    I was hoping that I could configure Distributed Resilient Network Interconnect (DRNI) on the two TOR switches in order to have device and link level resiliency. Then I can configure an LACP pair between the servers and the TOR switches.

    However, when I read the Layer 2 configuration guide from https://support.hpe.com/hpesc/public/docDisplay?docId=a00129122en_us

    there is a mention under the DRNI section: 'DRNI is an HPE proprietary protocol. You cannot use DR interfaces for communicating with third-party devices.'

    I do not understand what this means? Would it imply that the downstream devices connected to HPE TOR switches cannot be non-HPE?

    Thanks!



  • 2.  RE: DRNI compatibility with non-HPE devices

    Posted Jul 21, 2025 05:30 PM
    Hi friend.
    This does not imply that downstream devices connected to the DRNI-configured TOR switches cannot be non-HPE.
     
    Here's how it works in your scenario with Nutanix servers:
     
    DRNI between HPE Switches: DRNI creates a logical, resilient link aggregation between your two HPE 5980 TOR switches. From the perspective of the downstream devices, these two switches appear as a single logical entity.
     
    LACP with Nutanix Servers: The interfaces presented by the DRNI-enabled switches to your Nutanix servers will still support standard protocols like LACP (Link Aggregation Control Protocol). LACP is an industry-standard protocol for link aggregation.
     
    Third-Party Compatibility: Your Nutanix servers, being third-party devices, will use LACP to form an aggregated link with the logical interface presented by the DRNI-configured HPE switches. The servers don't need to understand DRNI; they only need to understand LACP, which is what the HPE switches will present.
     
    Therefore, you can configure DRNI on your two HPE TOR switches for device and link-level resiliency, and then configure an LACP pair between your Nutanix servers and the TOR switches. The Nutanix servers will see a standard LACP port channel, and the proprietary nature of DRNI is contained within the HPE switch pair.



  • 3.  RE: DRNI compatibility with non-HPE devices

    Posted Jul 22, 2025 05:40 AM

    Hi, out of curiosity, is there a particular reason (or a set of reasons) for not considering the IRF approach instead of the DRNI approach (especially if your ToR design will use - at maximum - only two HPE 5980 Switches)?




  • 4.  RE: DRNI compatibility with non-HPE devices

    Posted Jul 22, 2025 06:19 AM

    Hi, I understand that IRF will incur a longer downtime than DRNI during ISSU upgrades? We are considering DRNI as we would like to ensure high availability and resiliency as much as possible as well as to avoid any single points of failure.

    Could you advise on whether IRF has benefits to DRNI in a practical implementation as I personally have not observed IRF and DRNI in practice.




  • 5.  RE: DRNI compatibility with non-HPE devices

    Posted Jul 22, 2025 11:35 AM

    Hi, your asked "I understand that IRF will incur a longer downtime than DRNI during ISSU upgrades?" but the reality is that performing a Compatible ISSU upgrade procedure on a IRF stack with dual homed servers correctly connected to it (generally with LAG with LACP), was an engineered procedure - on the contrary - to produce no downtime at all (ISSU indeed means In-Service Software Upgrade). A different case is when an Incompatible ISSU upgrade procedure is started on the IRF Stack (jumping between major releases, generally) because there is always an IRF member active to which dual homed servers are actively connected to.




  • 6.  RE: DRNI compatibility with non-HPE devices

    Posted Jul 23, 2025 11:45 AM

    Hi, I add this document that can shed some light on the IRF and this one to show how to convert an IRF Stack to a DRNI Cluster (if needed). Cheers.