We have five ESXi hosts (VMware 6.7u1) cluster with vCenter 6.7 u1. We have run this configuration for about one year without problems.
Now it is not possible to migrate VMs to two of these five hosts by using vMotion. We get the error message: "The vMotion failed because the destination host did not receive data from the source host on the vMotion network.." when we try to migrate VMs to these two ESXi hosts.
We have a dedicated 40Gb/s NIC in every ESXi host for vMotion and all five ESXi hosts are configured with a VMkernel adapter and added the the same vMotion distributed switch with uplink..
The vMotion NICs on these servers are all on-line/connected in vCenter.
I have done the following:
I have also tested the VMkernel connectivity by using #vmkping from Putty.
From the two ESXi hosts I can ping the local VMkernal vMotion IP-address, but cannot reach the other ESXi hosts vMotion IP-addresses by using vmkping.
I am looking forward to hear from you to find a solution to solve this vMotion problem.
Thanks in advance!
Regards,
Audun
Welcome to the Community,
just to rule this out. Can you please double-check that the physical switch ports are all in the same VLAN, or - in case of tagged/802.1Q ports - that the VLAN ID on the vMotion VMkernel ports are set as expected? If the VLAN interface on the physical switch has an IP address assigned, check whether you can ping this one.
Also ensure that vMotion is only tagged on these portgroups (which of course should block a ping).
In case you 've configured Jumbo Frames on the VMkernel network, ensure that it's enable end-to-end.
André
Thank you for your response and good suggestions, Andrè.
We use VLAN ID 66 on these five vMotion kernel adapters and on these five switchports as well.
vMotion is only tagged on these portgroups.
MTU size is set to 1500
The strange thing is that this vMotion network has worked fine for about one year with no poblems and we have not done any changes.
So if you have any other suggestions or setting we should check, I would be very happy.
Thanks in advance!
/Audun
... and we have not done any changes.
Maybe not intentionally, but at least it's worth to double-check all settings to rule out accidental changes.
So please don't mind me asking again. Can you confirm that VLAN 66 is only either set on the physical switch ports, or on the VMKernel ports?
When you tested the connection via vmkping, did you specify the vmk to use in the command line options?
André
Thank you for follow up, Andrè
Of course I will double-check all settings you mentioned here on Monday.
I enabled SSH on the ESXi hosts and used Putty to login to the ESXi host and just from the command prompt # vmkping xx.xx.xx.xx (vMotion IP-address on the local and remote ESXi host). I got reply from the local host, but not from the Remote hosts.
From the ESXi where vMotion works, I got reply from the local host and the remote host vMotion IP-address.
Should I use any command line options?
/Audun
Although this may not change things in your case, testing VMKernel ports should be done by specifying the vmk.
e.g. vmkping -I vmk3 <IP-Address>
(that's an upper-case "i")
André
Thanks for your vmkping command line option e.g. I will use this on monday.
So you think our vMotion problem on these two ESXi hosts are due to a networking problem, VLAN ID on switchports, VLAN ID on kernel adapters or switchconfig and not a corrupted vCenter config on these two ESXi hosts?
Regards,
Audun
Well, I'm actually just guessing, because I don't have any details, or log files.
... and not a corrupted vCenter config on these two ESXi hosts
At this point vCenter is not involved. It's a configuration on the hosts themselves. It's just the VMKernel port group that matters for network connectivity.
André
Thanks for clarifying information. We use distributed switch for vMotion. Can it cause problems?
/Audun
Not, basically not. Only if there's a configuration issue with it.
André