Question regarding LACP with VLAN creates

[JGPSHNTAP]

Ok, so I need some clarification.

I'm currently running 8.1.2p2 and I'm wondering where i'm going wrong

My rc file looks like

# Network setup for 10 Gbit connection

ifgrp create lacp vif10g-a -b ip e1a e1b

vlan create vif10g-a 2508

ifconfig e1a flowcontrol send

ifconfig e1b flowcontrol send

ifconfig vif10g-a-2508 10.10.255.100 netmask 255.255.255.0 mtusize 9000 partner vif10g-b-2508

ifconfig vif10g-a-2508 nfo

route add default 10.10.255.1

routed on

When I fail the cluster over to the partner node, we get this message

hostname*> Replaying takeover WAFL log

ifgrp: vif10g-a is not mapped to a local ifgrp

add net default: gateway 10.10.255.1

And then I got to partner mode and type ifgrp status

"No configured ifgrps present"

But if I do ifconfig -a

i see the interface come up  (see below)

vif10g-a-2508: flags=0x2bee863<UP,BROADCAST,RUNNING,MULTICAST,MULTIHOST,PARTNER_UP,TCPCKSUM> mtu 9000

        takeover mode (vif10g-b-2508)

        ether 02:a0:98:31:dd:62 (Enabled interface group)

hostnamea/hostnameb> vif status

ifgrp: command "vif" is deprecated in favor of command "ifgrp"

No configured ifgrps present

vif10a-g-2508 and I can ping it

Is this normal behavior?

Shouldn't we see the vif come up? 

[dcornely1]

Can you provide the rc file from the partner node as well?  This way we can see what the network config is there to help with troubleshooting.

Also, I suggest you place your flowcontrol commands before your ifgrp create command (and if you're using VMware NFS datastores off this array then flowcontrol should be none) as it operates on the underlying interfaces not the ifgrp itself so you might as well set that first.  At least that's what I have in my environment and had no issues.

This may also help:

https://forums.netapp.com/thread/14286

[JGPSHNTAP]

Ok, the partner rc looks like

# Network setup for 10 Gbit connection

ifgrp create lacp vif10g-b -b ip e1a e1b

vlan create vif10g-b 2508

ifconfig e1a flowcontrol send

ifconfig e1b flowcontrol send

ifconfig vif10g-b-2508 10.10.255.101 netmask 255.255.255.0 mtusize 9000 partner vif10g-a-2508

ifconfig vif10g-b-2508 nfo

route add default 10.10.255.1

routed on

Yeah, i've been meaning to take off flow control

[AGUMADAVALLI]

Hi there,

I have provided the sample of how a rc and hosts files look below:

The best practice is to specify all the ips in the hosts files then the rc file.

RC file:

hostname __________

ifgrp create lacp vif10g-a -b ip e1a e1b

vlan create vif10g-a 2508

#Remove the flowcontrol send 2 lines

ifconfig vif10g-a-2508 `hostname`-vif10g-a-2508 netmask 255.255.255.0 mtusize 9000 partner  vif10g-b-2508

#No need of NFO line too

route add default 10.10.255.1

routed on

HOSTS file:

<IP>    <hostname>-vif10g-a-2508

Moreover, while you are trying for takeover specify the option -n which clear the wafl log errors.

Thank you

AK G

[JGPSHNTAP]

Couple of questions..

Why is this best practice -  `hostname`-vif10g-a-2508?

Also, why did you pull out NFO?

[AGUMADAVALLI]

NFO is not needed from 8.X as it is automatically enabled via the

options cf.takeover.on_network_interface_failure on
options cf.takeover.on_network_interface_failure.policy all_nics

thank you,

AK G

[JGPSHNTAP]

AK G -

Ok, i'm still a little confused here.. In partner mode when the cluster is failed over, should you not see any ifgrps?

[aborzenkov]

You still need to mark interface that has to trigger failover with "ifconfig ... nfo"