Comware

 View Only
  • 1.  HP 7500, Nutanix, GBP or insufficient bandwidth:

    Posted Jan 20, 2021 04:15 AM

    A customer have Nutanix AHV cluster (4 nœuds), connected to HP 7500 Switch Series, (2 x MPU JH209A and 2 x LPU JF290A), with firmware, 7.1.045, Release 7169P01. Using DAC HPE JD097C to connected LPU JF290A  to Nutanix.

     

    There is message appear on ports: Packets dropped due to full GBP or insufficient bandwidth: 1914

     

    Can you help me solve this problem?



  • 2.  RE: HP 7500, Nutanix, GBP or insufficient bandwidth:

    Posted Jan 20, 2021 04:58 AM

    Hello,

    Can you enable 'burst mode'?

    <Sysname> system-view
    [Sysname] burst-mode enable

    Thanks!



  • 3.  RE: HP 7500, Nutanix, GBP or insufficient bandwidth:

    Posted Jan 22, 2021 05:11 AM

    Thanks

    Now the  software version doesn' t supported the burst-mode, the customer is going to update the switch with the latest version.

    Hakim



  • 4.  RE: HP 7500, Nutanix, GBP or insufficient bandwidth:

    Posted Sep 07, 2021 09:12 AM

    But, remeber some tengigabit cards don't support "burst mode".



  • 5.  RE: HP 7500, Nutanix, GBP or insufficient bandwidth:

    Posted Sep 07, 2021 09:39 AM

    Hello @JUANTRON ,

     

    Yes, it is possibe.

    Can you share module product number 'JXXXXX'

    Thanks!



  • 6.  RE: HP 7500, Nutanix, GBP or insufficient bandwidth:

    Posted Sep 08, 2021 11:46 AM

    Hello. In my case, I have two HP 10508 chassis in irf cluster mode.

    I can't apply "burst mode" feature in the following cards:

    HP A10500 16-port 10-GbE SFP+ SC Module JC628A

    HPE 10500 16p 1/10GbE SFP+ SF Mod JH193A

    But "burst mode" is applied in the following cards:

    HP A10500 48-port Gig-T SE Module JC618A

    HP 10500 16p GbE SFP/8p GbE Cmbo SE Mod JC763A

     

     

     



  • 7.  RE: HP 7500, Nutanix, GBP or insufficient bandwidth:

    Posted Jan 22, 2021 04:49 AM

    Hi AbdelHakim!

     

    That counter shows you packets that are dropped because the buffer is used up or the bandwidth is insufficient. While enabling the burst mode is indeed the recommended action, you need to understand that there are situations when you will see outgoing dropped packets despite all efforts. For example, when 2 clients connected on 10G interface each simultaneously send traffic that goes out to a server over single 10G interface we have oversubscription situation and if the LPU where outgoing interface resides doesn't have deep buffers (most don't) outgoing buffers on that interface will get overflowed quite fast and LPU won't have any choice, but start dropping excessive traffic. "burst-mode enable" will try to automatically arrange buffer space to minimize such impact, but it can't do magic if resources are not enough to handle excessive traffic.

     

    BTW, TCP/IP has been designed to handle some traffic loss, but if you have requirements to avoid drops for some critical traffic or for storage traffic, you need to use QoS. It can help you to have no drops for certain types of traffic by dropping less important packets.