Cabling 3 nodes to a SAN through a Switch

[garethEds]

Afternoon All,

 

My journey to improve our HyperV setup continues after being setup a long time ago by @rich_tech. I'm in the process of adding a third node but before I do, I was wondering what the ideal way of cabling this all together is?

 

Two of the servers (HyperV1 and V2) have 6 network ports. The new server has 8 network ports.

 

The SAN is a P2000 G3 and the switch is a Cisco 200

 

Two notes is working well, but as I learn what I am doing I'm looking for improvements. So any wiring diagrams, advice etc would be gratefully taken on board.

 

Thanks

 

Gareth

[Steve21]

I mean it depends how you have the servers setup really. But the "standard-ish" way of doing it is split each "role" across 2 cards and 2 switches and 2 SAN modules.

 

e.g. Management Team Goes to Card 1 Port 1, and Card 2 Port 1.

Card 1 Port 1 goes to Switch 1 Port 1, Card 2 Port 1 goes to Switch 2 Port 1

Then SAN Module 1 Port 1 goes to Switch 1, and SAN Module 2 Port 1 to Switch 2

etc etc

 

Gives you as many failover points as possible should any one card/module/switch die

 

Steve

Edited May 11, 2018 by Steve21

[garethEds]

I mean it depends how you have the servers setup really. But the "standard-ish" way of doing it is split each "role" across 2 cards and 2 switches and 2 SAN modules.

 

e.g. Management Team Goes to Card 1 Port 1, and Card 2 Port 1.

Card 1 Port 1 goes to Switch 1 Port 1, Card 2 Port 1 goes to Switch 2 Port 1

Then SAN Module 1 Port 1 goes to Switch 1, and SAN Module 2 Port 1 to Switch 2

etc etc

 

Gives you as many failover points as possible should any one card/module/switch die

 

Steve

 

Thanks Steve,

 

Would this sort of information be on the web somewhere? With diagrams etc?

 

Gareth

[Steve21]

At a basic level it's like this:

 

 

So basically no matter what fails it has another connection. Repeat for each Team/Server etc

 

Steve

[garethEds]

At a basic level it's like this:

 

[ATTACH=CONFIG]48915[/ATTACH]

 

So basically no matter what fails it has another connection. Repeat for each Team/Server etc

 

Steve

 

Oooo - thanks Steve. Never thought of using two switches before. I only have one. I assume both switches would then link back to my main central backbone switch.

 

We have both our servers linked together too for failover. I think I may need to re-do ours. Is this how yours is set up?

 

Gareth

[Steve21]

Yep else the whole redundancy aspect is reliant on a single switch to fail if you have just a single switch there's not much you can do to stop a major failure if it breaks.

 

Well that's the storage side, generally "best practice" SAN->Host connections should have their own switches.

 

But yes it's the same principle, for example the HOST->Network side, is just two connections into our core switches (Which are stacked), So we have 2 connections in per host for Management, and 2 for VM traffic etc. (Those obviously don't need to go into the SAN switch if you have separate ones).

 

So effectively even if 1 SAN module, 1 Storage Switch, 1 HOST Nic and 1 Core fails you'd still have connection (Hopefully never would happen but yeah!)

 

Steve

[garethEds]

Thank @Steve21 - I'm going to draw this up similar to your diagram. What about IP addressing - I'll get to that next week.

 

Gareth

[Steve21]

Excuse my hastly painted example and obviously it's not just one link, but one per item e.g. Management, Traffic, Replication etc etc

 

but:

 

 

If that makes sense, and then per host you split it per card too

 

Steve

[Steve21]

Thank @Steve21 - I'm going to draw this up similar to your diagram. What about IP addressing - I'll get to that next week.

 

Gareth

 

That site depends if you're using multiple switches I guess.

 

We have our iscsi/storage on the 192s and our main networking/traffic on the 10s as those switches don't connection. But basically shouldn't really matter much as long as you keep it standard to make life easier when trouble shooting

 

Steve