Network card speed issue

[blu4]

Okay, I will try to explain this the best way I can so bear with me...

 

We have two servers, VH01 and VH02.

 

VH02 is connected to VH01 with a 10Gbps fibre link (Intel NIC)

VH01 is connected to the core switch with a cat6 1Gbps link (Emulex NIC is 10gbps but the core switch is 1gbps, so the connection is only 1Gbps)

 

The fibre link (Intel NIC) on VH01 (that connects to VH02) is bridged with the 10Gbps Emulex NIC so that VH02 can access the network. I have done it this way because the Virtual Machine replication traffic ideally should go over the 10Gbps NIC to avoid hogging the main network NIC. Now, the problem- when I copy from VH01 to VH02 (connected with a 10Gbps fibre) I only max at 1Gbps even though both adapters report 10Gbps and the bridge reports 20Gbps. Picture added for even less confusion.

 

 

tl;dr servers connected with 10Gbps but files copy at 1Gbps

Edited December 3, 2015 by blu4

[Sagima]

Have you tried setting the port speed on all the 10g connection to 10g and not leaving it at auto (or checking it's not actually 1).

[ITGURU]

Surely if your bridging a 1GB link and 10GB link, the max speed will always be the slowest link ?

[blu4]

Yea, just realised that. The vSwitch still thinks it's a 10Gbps however the link speed is 1gbps. I need to rethink my strategy.

[Steve21]

I need to rethink my strategy.

 

What's the point of the second server? as if it's anything more than "just" replication, you'll be having both servers going over a 1 gig link even if it did work. And then if the main one dies it'd lose connection anyway

 

Any reason the second server isn't connected directly to the network? So you wouldn't need to bridge?

 

Steve

[blu4]

What's the point of the second server? as if it's anything more than "just" replication, you'll be having both servers going over a 1 gig link even if it did work. And then if the main one dies it'd lose connection anyway

 

Any reason the second server isn't connected directly to the network? So you wouldn't need to bridge?

 

Steve

 

Unless it's a hardware issue, the second server is a failover with replicated virtual machines. The reason why I put it on just the 10Gbps link is that it would make the replication faster and it's a 10Gbps link, so why not use it anyway. We can't afford to get a 10Gbps switch at the moment. You made a very valid point though. I completely disregarded this fact.

Edited December 3, 2015 by blu4

[john]

Unless it's a hardware issue, the second server is a failover with replicated virtual machines. The reason why I put it on just the 10Gbps link is that it would make the replication faster and it's a 10Gbps link, so why not use it anyway. We can't afford to get a 10Gbps switch at the moment. You made a very valid point though. I completely disregarded this fact.

 

But with your setup you describe should host 1 die your replicated VMs to bring up won't come up and be useful as they have no network access.

 

Both hosts need to be setup identically to give you the HA you are suggesting you want and then you need to set replication to go over a VLAN / those NICs with fibre if they are directly connected to each other.

[Steve21]

Personally I'd ditch the bridge. Setup replication just to use the 10Gb link, and then put the second server directly into the network. That way both have access to the network (even if one dies), rep happens at 10gb and you have no issues

 

Also means you can run VMs split across both HOSTs and replicate both ways, to reduce bandwidth to a single machine etc

 

Steve

[blu4]

But with your setup you describe should host 1 die your replicated VMs to bring up won't come up and be useful as they have no network access.

 

Both hosts need to be setup identically to give you the HA you are suggesting you want and then you need to set replication to go over a VLAN / those NICs with fibre if they are directly connected to each other.

 

thanks for the tip!

 

Personally I'd ditch the bridge. Setup replication just to use the 10Gb link, and then put the second server directly into the network. That way both have access to the network (even if one dies), rep happens at 10gb and you have no issues

 

Also means you can run VMs split across both HOSTs and replicate both ways, to reduce bandwidth to a single machine etc

 

Steve

 

Great words of advice, thanks a lot. It's my first time doing this sort of project. Appreciate the help.