c7000 and GbE2c setup

[Robert Marek]

Hello, I’m looking for help with setting up c7000 and switches. My knowledge on this is limited.

This is what I have:
C7000 enclosure
4 blades with two mezzanine cards – one with dual NIC and on with quad NIC – total 8 NIC’s
8 GbE2c L2/L3 switches
Left Hand P4500 (3 nodes)

I have created VLAN on Switch 1 for 4 blades and 3nodes. All switches are interconnected internally with cross link ports, and uplink ports to connect switches vertically. This way all switches are connected together. The problem is if I lose Switch1 then I lose everything. It makes sense since VLAN is set on Switch1.
I would like to have a setup where no matter what switch fails my connections are up. I wonder if this is possible without creating VLAN on my core switch 5406zl. Right now my core switch is only passing traffic between my main network and VLAN on GbE2c switches.

Thanks in advance for any suggestions.

[Mike Schreiner]

have you gotten anywhere on this yet?

[Robert Marek]

Only what I posted. I'm wondering what would be the best setup for what I have from networking perspective. SAN is accessible and my ESXi host s are fine but this need to be bullet proof setup.

[Matt Palmer_2]

Hi,

I used spanning tree groups and LACP along with MSTP to make my setup fully redundant. do you have spanning tree in place on these switches or any higher up in your core network??

regards

Matt

[Robert Marek]

I do have spanning tree but I have no VLAN up on higher switch, in this case 5406zl. If I try contain VLAN only to switches in chasis it will never be fully redundant because if I loose first switch where VLAN is created I loose everything.

There only way to do this (in my opinion) is to setup VLAN higher up - on Procurve 5406zl and then make redundant setup on blade switches and if uplinked correctly I could loose 2 switches at the same and still have everything accessible... if I'm correct....Again, I have no prior experience with this so I may be totaly wrong.

[Matt Palmer_2]

Hi,

how I approached this was to break the interlinks on ports 17-18, and treat each switch as a separate entity.

then I created an aggregate on port 21-22 for 1 network 23-24 for other network, then used priority higher up in the network to manage failover if 1 or more cable was disconnected.

each switch has the same VLAN configuration on it, so within the OS (linux) I use active-backup bonding to cover myself if I loose an entire switch. I also used 802.1q in the os to tag the packets for each VLAN.

I can post up my switch config if you like, but you will have to tailor it a bit to your requirements, as I know nothing of your core network,etc.

regards

Matt

[Robert Marek]

Matt, I tried to keep my core network out of this setup and only pass trafic between blade switches (one VLAN) and core network.
If I understand you correctly this is not possible.

Ports 17-18 are disabled by default. I had to enable them to have access from switch A tp switch B.

[Matt Palmer_2]

Hi,

ok, not sure how to proceed with that one then. One other thing not sure if youve looked at is the bladeconnect forum. Its a pretty good resource for config and setup issues. you could try there - http://www.bladenetwork.net/

regards

Matt

[Antonio Milanese]

Hello Robert,

as Matt said a better solution is to configure MSTP and create a mesh topology within the encloser and use multiple(say 2)uplinks to your core switch.

>Matt, I tried to keep my core network out of
why ? insufficent modules/ports?
u need 6 port:
2x LACP (2x1Gbe x module) connected to two vertically/horizontally differents GbE2c (bay1/bay8 f.e.)

1x (1x1Gbe x module) connected to P4500 storage gateway with ALB loadbalancing

GbE2c are nothing more than regular L2/L3 switchs so MSTP is a must have with multiple redundant links

>blade switches (one VLAN) and core network
it's advisable to have a different dedicated VLAN for storage for at least 3 different reasons:

1) broadcast isolation
2) security
3) loadbalancing using MSTP instances

my suggestion is to create
2 MSTP instances, a storage VLAN mapped on all pNICs used as uplinks for ESX iSCSI portgrop/vswitch:

this way you can create 2 different paths within the encloser (i.e. balanced on left,righ bays) and on uplink pipes reserving effectively more bandwidth for your storage traffic.

Regards,

Antonio

[Robert Marek]

Thanks Antonio,

I have one VLAN (on switch 1) for storage and ESX hosts to seperate traffic from main network. Switch 2 is connected on ports 17-18. Both of them are uplinked with core switch on default VLAN 1. If I power down Switch 1 VLAN is down and nothing is accessible from my network. This make sense sincethis Switch is setup with this VLAN IP.

I was hoping to be able to setup everything within the chasis without pushing any traffic through core switch.

I guess I'll have to setup VLAN's on my core switch...

Right now I have NIC#1 from each Lefthand node connected to switch 1 and NIC#2 connected to switch 2.