I've been banging my head against this wall for long enough, so I am hoping someone can shed some light on where things have gone so wrong.
I recently deployed a UCS environment consisting of the following:
- 5108 w/ 5xB200 M4 blades
- 2 x 6248 (2.2 (5a))
- EMC VNXe 5200 (direct connected iSCSI)
- vSphere 6u1
- ESXi 6u1a (Cisco custom image) with KB2137546 patch and enic 2.1.2.71
The FIs uplink to a pair of 3850s (stacked).
The hosts and VMs have been exhibiting intermittent connectivity issues and poor performance. Consistently, if I reboot the host, ESXi management connectivity is lost until I reset the management network within ESXi.
From the vSphere side, I have a pair of vmnics on the management VLAN as well as other production VLANs, and a pair of vmnics for iSCSI. One standard vSwitch for mgmt. and production traffic (teaming based on Port ID), and a second standard vSwitch for iSCSI traffic (no teaming).
I have tried a number of things to troubleshoot and resolve the issue. Recently, I discovered the critical issue with 2.2(5a) and poor performance on B200 M4 blades. I upgraded the software to 2.2(5b) this past weekend with no resolution.
At this point, if I reboot a host with two vmnics on the mgmt. vSwitch, the host will have NO mgmt. connectivity. If I reset the management network, it responds to pings. If I remove a vmnic and reboot, everything reboots happy, albeit without redundancy.
If I leave the two vmnics on the mgmt. vSwitch and reboot, no connectivity. But if I disable one of the NICs in UCS manager, it starts to respond to pings. As soon as I re-enable the NIC in UCSM, pings are dropped.
So, I'm at a loss. Any insight? Thanks
