Fas 2020 for NFS datastore vmware

[reinhartman]

Hi all

The last day i am setting up a fas2020 (dual controller) for our vsphere 4.1 farm.

The filer will only be used for nfs datastores for the vm's.

The company i setting this up for doenst have switches that supports cross switch trunking / etherchannel. So i came with the following config:

I use a single vif on eatch controller. Both nics are connected to an other switch. (I use favor to select the nic i favor for each controller, on this way i can split traffic accros the switches)

I also use vlans, for the simple reason i use 2 subnets for vmware (1 per vmkernel) and a 3th for management.

Controller 1:

hostname brena01
vif create single vifna2 e0b e0a
vif favor e0a
vlan create vifna1 1
vlan add vifna1 12
ifconfig vifna1-1 192.6.1.50 netmask 255.255.255.0 mtusize 1500 partner 192.6.1.51
ifconfig vifna1-12 10.0.1.3 netmask 255.255.255.0 mtusize 1500 -wins partner 10.0.2.3
mediatype auto partner vifna2
route add default 192.6.1.254
options dns.enable on
options nis.enable off
savecore

Controller 2:

hostname brena02
vif create single vifna2 e0b e0a
vif favor e0b
vlan create vifna2 1
vlan add vifna2 13
ifconfig vifna2-1 192.6.1.51 netmask 255.255.255.0 mtusize 1500 partner 192.6.1.50
ifconfig vifna2-13 10.0.2.3 netmask 255.255.255.0 mtusize 1500 -wins partner 10.0.1.3
mediatype auto partner vifna1
route add default 192.6.1.254
options dns.enable on
options nis.enable off
savecore

Now, i have it all running and it works, but here comes the problem.
When i do a takeover of controller 2:

- I am able to ping both addresses 192.6.1.50 and 192.6.1.51 ==> I conclude that my partner / vlan config is oke

- I am not able to ping 10.0.2.3 ==> so vmware loses is access to the nfs datastore on this controller

I am probaly doing something wrong with the 2 subnets?

The ports of controller 1 and 2 are all tagged in VLAN 1-12-13

Would be very nice if someone could say if its correct what i am doing, or is it totaly wrong.

Many thanks in advance

Reinhart De Bock

Message was edited by: reinhartman

[ademuynck]

Hie,

Your network config is :

ifconfig vifna1-1 192.6.1.50 netmask 255.255.255.0 mtusize 1500 partner 192.6.1.51
ifconfig vifna1-12 10.0.1.3 netmask 255.255.255.0 mtusize 1500 -wins partner 10.0.2.3

So for your first interface, all is allright. both ip are on the same subnet.

So for your second one, both ip are on two different subnet.

Have you try to put in the same network and then test your takeover ?

In fact, for your vlan 12, you have to set this :

ifconfig vifna1-12 10.0.1.3 netmask 255.255.255.0 mtusize 1500 -wins partner 10.0.1.4

I think your problem is here.

So your final setup look like this :

ifconfig vifna1-1 192.6.1.50 netmask 255.255.255.0 mtusize 1500 partner 192.6.1.51

ifconfig vifna1-12 10.0.1.3 netmask 255.255.255.0 mtusize 1500 -wins partner 10.0.1.4

[reinhartman]

Hey

Thanks for the reply.

Placing vlan 12 and 13 in the same subet is not possible for us.
In fact, we need to have both controllers in 2 different subnets so i can make 2 vm kernelports  and make an datastore on each controllers.

The vm kernelports are each in one of the above subnets and each go over an other switch to a different controller.
With this setup i have 1 active path to each controller over a different switch. (i have no spof)

This is working all fine, just in case of an takeover i lose connectivity to one datastore.

Reinhart

[ademuynck]

You're welcome

So ok.

And if you create an alias for 10.0.0.13 (ex:10.0.2.13) and an alias for 10.0.2.3 (ex: 10.0.0.3)

Not sure it will works ... But you can try this ...

[reinhartman]

Hey

I did indeed test with an alias this afternoon.
When i set the vifna2-13 address (10.0.2.3) as a alias on vifna1-12 it works again.

Ofc, this is not a proper solution as it has to be done manualy. ( in proper running without takeover) i would have duplicate addresses.

Vmware connects to a single address per nfs store, so i need to stick with that address

Thanks for the brainstorming!

Reinhart