Virtual Machine redundancy

[scale21]

Hopefully this is the correct forum as i am using esx 4.1

Here we go:

I have x number of hosts. (ill be using 2 to explain but know i have more)

I have 1 vswitch that runs all of our VMs bound to 2 phyiscal nics. These 2 nics are plugged into 1 physical switch

WE have many port groups broken out into many different port groups on different vlans to create separate little networks.

We have as many as 12 port groups on this vswitch bound to both pnics using a route based hash algorithm for the load balancing.

The physical switch:

each host has 2 connections to this unit.

For example.....host 1.....pnic1 and pnic2 are plugged into hp switch port 1 and 2.

the switch port 1 and 2 are trunked together and added to their respective lans.

we are then able to access each vlan respectively to use the small networks created over these 2 ports and have a small level of redudnancy.

probelm:

There is only 1 physical switch.

IF that dies....i could have 30 ports trunked.....who cares...im still dead in the water.

i know i need 2 physical switches for my front end to have any sort of redundancy.

That said.....

Im thinking of going with 2 hp 2910-48g units.

My questions are with hookup

If i go down this path, i will want to have 2 uplinks from switch 1 to switch 2 for redundancy. Will spanning tree allow for this to happen or what prevents loops in this type of setup?

ScenerioA would be 2 hp procurve 2910 switches with 2 1gb uplinks.....lets say port 47 and port 48 on each switch will be uplinked to the other.

ScenerioB would be to look at a 2 10GB modules that i could put into the rear of the units and run 2 seperate 10gbE cables from one switch to the other.

Again....will spanning tree sort out the loops here? Is there any special configuration that needs to happen in a situation like this?

Lets say i get that figured out and the switches are now set.

As mentioned above ive got host1 pnic1 and pnic2 are plugged into switch port 1 and 2. Ports 1 and 2 are trunked together.

I would think that the trunk then can be removed and host1 ... pnic1 gets plugged into port 1 on switch 1

pnic2 gets plugged into port 1 on switch 2

does it sound like a proper plan to start my journey into redundant switching?

I have inherited this system so im trying to go in with a cautious approach.

[a_p_]

What I don't understand in your scenario is why you want to trunk the ports together with the default port based policy? With the 2910 switches I'd go ahead and configure them for e.g. RSTP and setup all the ports as tagged (i.e. no trunks). ESXi will assign each VM to only a single uplink at a time, so there's no reason to be afraid of networking issues. The only trunk which makes sense in this case is for the 10GBit module ("A1" port).

André

[scale21]

thanks.

yeah...im confused by the trunk as well. I am not sure why it was done. I plan to desolve one going to one of my hosts to see if there is any reason not to just desolve them all and run with 2 pnic connections without them trunked together (hp trunked that is).

My post is a bit confusing and long.

Bottom line is i want to introduce a second physical switch for my virtual machine nics to plug in to. I  also want redundancy.

So....

Switch1 -->pnic1 from host 1

Switch2 -->pnic2 from host 1

Run 2 uplinks between both switches for redudnancy sake.

tag all ports involved.

turn on spanning treee (which i believe is on by default in procurves world)

done?

seems simple.....i just dont want to miss anything.

[scale21]

I should be clear when i say trunk. I am talking about HPs trunk terminology where 2 switch ports act as one port. I know it might be something different in the cisco world.

I am not familar with RSTP.

I believe my switch is current using MSTP. That is what it shows when i do a "show spanning-tree config" in my procurve.

It also says STP Enabled: NO

This tells me that spanning tree isnt even on.....

Either way....STP is for preventing loops. If this is the case and i had 2 switches for redundancy, in theory isnt it only going to use one path until it fails at which time the other path becomes active?

If this is the case i can see why the previous person who set this up added the TRK ports on the switch. IT provides a single logical port across 2 physical ports. STP doesnt stumble over it during a port failure....and also....it appears to offer a bit of load balancing across switch ports.

I guess i am still back to my original question. How do i introduce a second switch and make it redundant.

Not sure if there is a way to TRK a single port from 2 different physical switches. That would do the trick.....however that seems impossible.