VM having internet issues + Bonding NIC Questions

[Naldinho]

Ok so I have unraid working with some VMs but I am having issues with the main VM's internet connection -- I have it as a bridge and it works except when it doesn't -- I'm getting a lot of webpages that just won't load and I'm also getting webpages that load but have images missing or sometimes they are just half text. When I run a speed test though I get close enough to my ISP speed 25/10 that I know it can sustain that.

 

In addition from my separate computer I am occasionally having issues with loading of the WebHui on Tower. It loads but slowly, sometimes with graphics missing, sometimes with just the frame but no data. This is also intermittent.

 

The other two VMs are headless and I typically have one or both open on my separate computer using RDP. They seem to work fine but I don't do any web browsing on them or any heavy internet use.

 

Everything is going through a budget 100M D-Link switch.

 

Obviously my first suspicion is that it is the switch which I was going to upgrade anyway but I figured I'd seek advice first.

 

1) Am I correct in assuming it is the switch or is this a known problem with Windows 10 VMs because the drivers are still new?

 

2) My mainboard has dual Giga NICs -- I am going to be researching the multiple options to bind them but would it be worth getting a switch with LACP to take advantage of some of the bonding options that need it?

 

3) Do I even need to bind the two NICs or will I be fine once I have a Giga switch?

 

4) Looking in the topics it seems that I can't passthrough the second NIC to the Windows 10 VM. Would I be better off getting a PCIe NIC and passing that though so that the main VM doesn't have to share the bridge mode NIC?

 

 

 

 

[JWMutant]

I can rule out it being a windows 10 Issue as I have a Windows 10 VM on my tower and works fine.

 

I did have connectivity issues which I solved by giving my VM a static ip in my router.

[f3dora]

1) Am I correct in assuming it is the switch or is this a known problem with Windows 10 VMs because the drivers are still new?

I can rule out it being a windows 10 Issue as I have a Windows 10 VM on my tower and works fine.

 

Can confirm, win10 vm with gpu passthrough and no connectivity issues

 

2) My mainboard has dual Giga NICs -- I am going to be researching the multiple options to bind them but would it be worth getting a switch with LACP to take advantage of some of the bonding options that need it?

 

Get a switch that supports 802.3ad Link Aggregation https://en.wikipedia.org/wiki/Link_aggregation

 

3) Do I even need to bind the two NICs or will I be fine once I have a Giga switch?

 

If you have multiple devices accessing your server or use your server to host a plex server or NAS and use it as a workstation or Gaming PC you should definitely use NIC bonding.

 

4) Looking in the topics it seems that I can't passthrough the second NIC to the Windows 10 VM. Would I be better off getting a PCIe NIC and passing that though so that the main VM doesn't have to share the bridge mode NIC?

 

No, if your main vm does not have enough bandwidth on the local network you should get a switch that supports bonding and bond the onboard nics together using 802.3ad.

[bonienl]

No, if your main vm does not have enough bandwidth on the local network you should get a switch that supports bonding and bond the onboard nics together using 802.3ad.

 

Linux supports different types of bonding modes, not all of them require a switch with specific support.

 

In this particular case I would recommend to go for a gigabit switch and replace the 2x100 for 1x1000 connection.

 

[Naldinho]

I did buy a switch that supports 802.3ad Link Aggregation and that seems to have solved the problem.

 

It did create a new issue to fix though -- one connection is at 1000 the other at 100.

 

Both NICs are connect at 1000 individually with the other disconnected.

 

If I connect both without bonding the first one physically connected negotiates at 1000 the second one connected negotiates at 100 -- so I can either have 1000/100 or 100/1000 depending on if a physically connect eth0 or eth1 first

 

If I use bonding then eth0 gets to be at 1000 and eth1 is at 100.

 

I don't know what I'm doing wrong or if this is simply a limitation of the S2600CP motherboard.

[bonienl]

What bonding mode did you configure ?

 

Try bonding modes 1, 5 or 6 (with the switch configured as regular ports).

 

[Naldinho]

I'm not in front of the machine but I used 802.3ad Link Aggregation which I think is 4?

 

I'll try the other modes to see if anything changes but even before I turned on bonding only one NIC was negotiating to 1000.

[bonienl]

I'm not in front of the machine but I used 802.3ad Link Aggregation which I think is 4?

 

I'll try the other modes to see if anything changes but even before I turned on bonding only one NIC was negotiating to 1000.

 

You are saying: if the NICs are connected one at the time, they each negotiate to 1000M, but once both NICs are connected at the same time, you end up with 1000M / 100M ?

 

[Naldinho]

You are saying: if the NICs are connected one at the time, they each negotiate to 1000M, but once both NICs are connected at the same time, you end up with 1000M / 100M ?

 

Correct. I can connect either NIC and it will be 1000M but as soon as I connect both the second one connected goes to 100M. This happens with or without link aggregation turned on.

[Naldinho]

Well I feel kind of stupid. Turns out it was just one of the CAT wires. Problem solved

[Naldinho]

And now today it is back to 1000M / 100M so not the cable after all. This is really strange.