When I ping the VMs client hosted by Esxi 6.7 U2, first requests sometimes timed out

[jmd75]

Hello,

I've an HP Proliant DL380G9, recently installed with HPE Gen9 Plus Custom Image for ESXi 6.7 U2.

The HP server has been updated witch recent firmwares.

3 Windows 2012 standard hosted, with 2 vswitch each using a network card connected to two different physical switches.

The network cards for the VMs are of VMXNET3 type.

Frequently, from any host - same Ip v4 subnetwork - the first ping requests (the first  and sometime more) fail with "request timed out" response. After that, all works fine. It occurs fot the 3 virtual machines of this Esxi host.

I don't think this is a hardware problem (same switches are used with similar environments except the VmWare version).

I've noticed too, with wireshark utility, on one of the VM client, some ICMP frames with the IPv4 address  of this client as souce and destination and a "destination unreachable (host unreachable)" message in the info column.

I've patched today the Esxi with the build 201906002 (june). But always this problem.

Any idea?

JM

moderator edit by wila: Discussion moved from Workstation Pro to ESXi

[Alex_Romeo]

Hi,

If you have free ports on the physical switch, try changing the ports that connect the Esxi that has the ping defect.

Alesandro Romeo

[jmd75]

Hello,

I've already connect to another port, on an other switch.

There is two VM's on a first vswitch connected to a first physical switch, a third VM on the second vswitch connected to a second physical switch.

But all the ping to these VMs have the same random time out.

(the VmWare tools have been updated too).

That's strange.

JM

[Alex_Romeo]

Hi,

some ideas can be taken from these articles.

VMware Knowledge Base

VMs spontaneously lose network connectivity - Networking - Discussions

Alessandro Romeo

[NathanosBlightc]

Please draw a schematic from your ESXi Networking (state of vSwitch and port group) or take a picture of its current configuration.

Did you ever check uplink ports on physical switch side? is there any specific security configuration like etherchannel?

And also tell me about your uplink failover policy on vswitch and state of uplinks (Active/Standby)

[jmd75]

Thank you for you advices.

This is a track that I had already exploited, but I followed the recommendations of this article for all network cards.
https://kb.vmware.com/s/article/1004089

(I've found this link, by following the different links you gave me).

This is maybe obvious but the switch ports are often with autonegotiation, and the default configuration of the Vmware NICs, is not "auto, Unlike older VmWare versions.
I set "auto" on all the ports of the VmWare host, the management port too (the speed was "100 Mbps", an now the speed is "Giga").

I've do some changements, to my networking too (the main ports of the VmWare host on the same switch).

With this changements  it seems better now. I'll see in the days to  come.

JM