Slow ESXi performance (HF40)

[preecy]

We recently puchased a full HPE solution consiting of a single HF40 storage unit, three DL360 Gen10+ servers and two Mellenox switches. As we're new to HCI/Nimble we also purchased the onsite HPE installation/setup service.

All the hardware was installed and confirmed by HPE engineers and they deployed a new vcentre v7 app for us. Before we begin to move our production vm's over from our old HP SAN setup, we wanted to doing some basic testing but found we were sometimes seeing slow performance/latency. We managed to track this down to one of the DL360 ESXi servers (ie: host01). For an example, we can vmotion vm's to host01 in seconds but when trying vmotioning vm's off, it takes 30+ minutes. By the way, this is a simple migration of compute only (not storage) from one host to another. We also noticed, on the Mellenox switches, the management nic for host01 is generating a lot of rx read errors. HPE Nimble support have double checked the swtich and ESXi settings and confirmed all is good (ie: mtu, flow control, etc). HPE have also replaced both dual 25gb Broadcom nics in host01 but this didnt resolve the issue. We tried upgrading the nic firmware to v219.0.17.0 but again this hasnt helped?

We're currently waiting on HPE to advise the next course of action, however, we could really do with a fix ASAP.

Anyone been in a simpler situation or offer some advise on where we go next?

[support_s]

System recommended content:

1. VMware ESXi 6.0 - Support Guides | VMware ESXi 6.0 - Deteriorated I/O Performance on the Local Data Store

2. HPE Helion CloudSystem 10.0: Performance and Scalability Solution with KVM and ESXi Hypervisors

Please click on "Thumbs Up/Kudo" icon to give a "Kudo".

Thank you for being a HPE valuable community member.

[Vipin_S]

Hello Preecy,
It looks like the issue revolves around the Mellanox switches and Vcenter configuration.

From Nimble storage side already the Nimble support team is involved, and they didn't find any issues. In that case lets wait for update from the HPE case you created for the issue. I will also check with my colleagues if anyone experienced similar issues and update you accordingly.