ESX and Netapp

VBN

VBN

VBN

Associate
Joined
30 Jun 2005
Posts
149
Location
Newbury
Following on from my post about separating out iSCSI for ESX and NFS etc I was wondering if anyone would care to comment on the following design....

My NetApp filer has 4 nics, assuming that I’m not going to route for switch failure I was thinking of setting up the following.....

Create a VIF on all four Network Nodes
Create 4 VLANS on VIF
999 – Regular Network on 10.10.20.X 255.255.252.0
1000 – iSCSI Network on 192.168.250.X 255.255.255.0
1001 NFS Network on 192.168.251.X 255.255.255.0
1002 iSCSI Network (for iscsi luns) 192.168.252.X 255.255.255.0

On ESX

vSwitch0 * 2 Nics
Service Console
Internal Production - used by vm's for regular network traffic

vSwitch1 * 2 Nics
VMKernel
VMotion – 192.168.250.X

vSwitch2 * 2 Nics
VMKernel
NFS – 192.168.251.X

vSwitch3 * 2 Nics
VMKernel
iSCSI luns (Used for SQL, Exchange data and NetApp Snap Manager Tools)

vSwitch4 * 2 Nics
DMZ Production - webservers network access etc

The Filer and ESX servers are connected via a Cisco2960G (layer 2 switch)

To be honest I’m not sure what would need to be set on the switch, trunk ports etc. Also any ESX additonal configs, would I need to tag the traffic on the ESX nics for the filer to see it or because they are all different subnets would be OK?

Would be grateful of any advice.
 
You'll need to tag the traffic on each vswitch, easy enough though.

Last time I spoke to Netapp and VMware though they were tacitly recommending NFS for datastores as not only is it very flexible features wise it was also a little quicker than iSCSI. May have changed since but certainly that was direct from the manufacturers when I last went out to stockley park to see them..
 
VBN said:
To be honest I’m not sure what would need to be set on the switch, trunk ports etc. Also any ESX additonal configs, would I need to tag the traffic on the ESX nics for the filer to see it or because they are all different subnets would be OK?

Would be grateful of any advice.
Click to expand...

does the switch support LACP? if so I'd do:

on the filer make the vif a multi that supports LACP

on esx change vswitch 0 & 4's loadbalancing to source/dest IP

on the switch
change the etherchannel load balancing setting to source/destination IP (not sure if this actually has an effect on layer 2 etherchannel)
create etherchannel0 interface as a trunk
bond the four switch interfaces the filer is cable to to the channel group with mode set to desired and set them as trunks
create another etherchannel for each vswitch you have in ESX
bond the switch interfaces the vswitch NICs are cable to to the associated channel group with mode set to on (ESX AFIK doesn't support LACP yet) and set them as trunks

we're just starting to look at iSCSI on our filer and NetApp advised us to have two iSCSI VLANs with a VLAN interface for each one, though both are on the same mvif. We'll be creating therefore two iSCSI VMKernals, one to go in each VLAN, I guess to give us segregated multipathing. We'll then statically tie each iSCSI VMkernal to an individual NIC with no standby adapters.

I'm not too hot on iSCSI so I'm not sure if I should set the NetApp iSCSI vifs as shared or not encase of a controller failure, so I've set them to be dedicated, not sure if thats right!!
 
oh and we're migrating our ESX datastores away from FCP to NFS, which is another learning curve. iSCSI is there only for our SME setup
 
You must log in or register to reply here.
Back
Top Bottom