weird ping results from both inside and outside proxmox network

[damon1]

Hi Guys,
I have a 4 node Proxmox setup each node has 5 or 6 Windows 10 VM's on it.
Everything works great except they cannot ping each other with any certainty.
I might get 0 of 4 replies or 4 of 4 or as shown below 2 of 4.
This is from a VM on node 2 to a VM on node 1




If I go to node 1 and ping the VM on node 2 I don't have a problem



Similarly, if I go to Node 2 and ping the VM on node 1 there is no problem



However,
If I go to the VM on Node 1 and ping the VM on node 2 then it will work (twice)



BUT if I use the same VM on Node 1 to ping another VM on Node 2 it will fail



The only non "out of the box" network addition is for Ceph, a dual 10GB network card. I.e. Network Config for node 2


Lastly,
If I use a physical PC connected to the same network then i CAN ping all of the nodes consistently.


If I use a physical PC connected to the same network then there are some VM's I can ping reliably, some that wont respond and some that will give me 1 of 4 replies.





If I localize to the physical PC (Windows 10) then I have tried
With Firewall / Without Firewall
Change CAT cable
Change switch
Change DNS to 8.8.8.8 and 8.8.4.4
Nearly, everything I can find on Google

From Node 1 I can ping all of the VM's consistently which reside on Node 1 etc...

Problem:
I want to connect to 192.168.0.47 from the Physical PC but it gives very poor connection via a TCP connection (port 3306) .
<Cannot attach anymore pictures but I get 75% packet loss>

I can use windows to browse the shared folders on the VM (and all VM's) with no problems

Any ideas anyone, any suggestion greatly appreciated?

Thanks Damon
PS - now a subscriber

 

[wolfgang]

Hi,

what nic type do you use as vNic?
And is you use virtio what is recommended what driver do you use (please the iso version).

 

[Stoiko Ivanov]

apart from what @wolfgang pointed out - also check for duplicate IPs/MACs in the network (would explain the 50% packet loss on ping and 75 on tcp)

 

[damon1]

Hi Guys,
thanks

On Node 2 this is the network configuration, the other nodes are the same



On each VM I use the same configuration


The MAC address is also on the router to obtain a Static IP.

the VIRTIO drivers are from win 0.1.173, which I believe is the most recent which I obtained from
https://docs.fedoraproject.org/en-U...tual-machines-using-virtio-drivers/index.html

would the drivers on "Latest virtio-win amd64 floppy" be the same? All my nodes are AMD processors.


in the windows Driver details I see this


If I cut back to the bare bones and use the router (EdgeRouter 6P), I can ping all the NODES fine and find that the VM's that would not connect or had limited connection to the Physical PC have the same problems.
I can ping the physical PC from the router as well with no problems.




Just wanted to eliminate the Physical PC as the problem.


I thought about a duplicated MAC address as well but the MAC->IP is handled by the router and according to its lists all the MAC's are different as are the IP address's

Sorry it isn't more informative.

thanks
Damon

 

[damon1]

Since you mentioned the Network Device's I added another Network Device to one of the "BAD" VM;s to test


which gave a new MAC and IP 192.168.0.136
when i tried to ping it there was no connection either

 

[wolfgang]

Try to disable the firewall on the vnic.
After this restart the VM.
If this does not help try with another switch.
Maybe the EdgeRouter 6P has a feature/bug that block the communication.