[PVE-User] Intermittent NIC Connectivity after Host Upgrade to PVE 4.15.18-43

JR Richardson jmr.richardson at gmail.com
Wed Jul 10 17:04:53 CEST 2019


Hi All,

I recently upgraded a host within a 4 Node Cluster from
pve-manager/5.2-5/eb24855a (running kernel: 4.15.18-1-pve) to
pve-manager/5.4-10/9603c337 (running kernel: 4.15.18-17-pve).

I had several VMs running on the host node, some Win2016 Servers and
Linux Servers. I migrated the VM off the host node, performed the
upgrade, them migrated the VMs back to the upgraded host. All seemed
to go well, next day I notices the Win2016 servers having intermittent
network connectivity, Linux servers running fine. Performing
continuous pings to Win2016 servers, the NIC were dropping for 45-60
seconds every hour or so, time was not consistent. I was using VitIO
for NICs.

I did windows update and this crashed the NICs, would not come up. I
downloaded the latest virtio-win-0.1.171.iso and updated drivers, This
did restore the NICs n the VMs but still have the network drops. I
switched the VM NICs to Intel E1000 and still have the network drops.

I don't see anything in the host node logs or the Win2016 VM logs
indicating why the network drops periodically. Could this be a PVE
kernel issue? I think this is more of a Win2016 issue, maybe the Intel
NIC driver, its using the embedded Windows driver from 2010. Should I
try downloading the latest driver from Intel?

Dell R710 host node NICs:
Ethernet controller: Broadcom Limited NetXtreme II BCM5709 Gigabit
Ethernet (rev 20)
Ethernet controller: Intel Corporation 82576 Gigabit Network Connection (rev 01)

The NIC on the host for VMs is Broadcom. But the Linux hosts do not
experience any drops so i don't think this is a host node or host
kernel issue but I'm not sure.

I appreciate any guidance.

Thanks.

JR
-- 
JR Richardson
Engineering for the Masses
Chasing the Azeotrope



More information about the pve-user mailing list