Mailing List Archive: Netapp: toasters

Network routing question

Index | Next | Previous | View Threaded

scl at sasha

Apr 16, 2009, 12:56 PM

Post #1 of 4 (1702 views)
Permalink

Network routing question

Does a host route override a network route on a netapp head?

I have an interface plugged into one switch
with IP 192.168.5.6 netmask 255.255.255.0 and

I have a second interface plugged into a different switch
with IP 192.168.5.7 netmask 255.255.255.0

One switch is "primary" and the other is "standby" for
redundancy.

There is obviously a routing ambiguity here. Which port
does the filer use to reach the 192.168.5 network? It
seems to have picked one port (fortunately on the 'active'
switch) which I see listed in the routing table. The other
port is not listed in any routes at all.

But I was wondering if I could add explicit host routes
to the IP addresses on the standby switch. I tried this
and it didn't work (ping failed). The destination IP
that I tried to ping was on another filer head. It occurred
to me that I probably need to set the corresponding route
on the other head, so I did that. The routes are listed in
the route table, but ping still does not work. I checked
the ARP table and neither head has the MAC address of the
other in its ARP table. So I suppose that despite the host
route, the ARP broadcast is going out the wrong interface.

I think we need separate subnets here but I'm just the
storage guy and did not configure the hosts that are plugged
into these two switches.

Things are working, all traffic is on the active switch. But
I sure don't see how things could ever fail over to the
standby switch without manual intervention.

Steve Losen scl [at] virginia phone: 434-924-0640

University of Virginia ITC Unix Support

tmacmd at gmail

Apr 16, 2009, 1:07 PM

Post #2 of 4 (1606 views)
Permalink

Re: Network routing question [In reply to]

what about....

vif create single vif-s0 e0a e0b
ifconfig vif-s0 `hostname`-vif-s0 netmask 255.255.255.0 mediatype auto
flowcontrol full
vif favor e0a

assume e0a and e0b are your interfaces.
Create a single mode vif (use only ever, one interface at a time)
Bring up the vif
Favor the favorite switch.

If that switch fails, the filer automatically moves the link to the
standby (e0b).
When e0a comes back, I believe it moves back to it...the favorite.

yes?

--tmac

RedHat Certified Engineer #804006984323821 (RHEL4)
RedHat Certified Engineer #805007643429572 (RHEL5)

Principal Consultant

On Thu, Apr 16, 2009 at 3:56 PM, Stephen C. Losen
<scl [at] sasha> wrote:
>
> Does a host route override a network route on a netapp head?
>
> I have an interface plugged into one switch
> with IP 192.168.5.6 netmask 255.255.255.0 and
>
> I have a second interface plugged into a different switch
> with IP 192.168.5.7 netmask 255.255.255.0
>
> One switch is "primary" and the other is "standby" for
> redundancy.
>
> There is obviously a routing ambiguity here. Which port
> does the filer use to reach the 192.168.5 network? It
> seems to have picked one port (fortunately on the 'active'
> switch) which I see listed in the routing table. The other
> port is not listed in any routes at all.
>
> But I was wondering if I could add explicit host routes
> to the IP addresses on the standby switch. I tried this
> and it didn't work (ping failed). The destination IP
> that I tried to ping was on another filer head. It occurred
> to me that I probably need to set the corresponding route
> on the other head, so I did that. The routes are listed in
> the route table, but ping still does not work. I checked
> the ARP table and neither head has the MAC address of the
> other in its ARP table. So I suppose that despite the host
> route, the ARP broadcast is going out the wrong interface.
>
> I think we need separate subnets here but I'm just the
> storage guy and did not configure the hosts that are plugged
> into these two switches.
>
> Things are working, all traffic is on the active switch. But
> I sure don't see how things could ever fail over to the
> standby switch without manual intervention.
>
> Steve Losen scl [at] virginia phone: 434-924-0640
>
> University of Virginia ITC Unix Support
>
>
>

dleeds at edmunds

Apr 16, 2009, 1:20 PM

Post #3 of 4 (1625 views)
Permalink

RE: Network routing question [In reply to]

if this is a redundant setup why would you not configure your switches with 802.3ad link aggregation and configure a
vif on your filer with those two interfaces?

how is your setup even redundant anyhow you would have to remount all the clients manually if the primary interface died or
the primary switch died.

________________________________________
From: owner-toasters [at] mathworks [owner-toasters [at] mathworks] On Behalf Of Stephen C. Losen [scl [at] sasha]
Sent: Thursday, April 16, 2009 12:56 PM
To: toasters [at] mathworks
Subject: Network routing question

Does a host route override a network route on a netapp head?

I have an interface plugged into one switch
with IP 192.168.5.6 netmask 255.255.255.0 and

I have a second interface plugged into a different switch
with IP 192.168.5.7 netmask 255.255.255.0

One switch is "primary" and the other is "standby" for
redundancy.

There is obviously a routing ambiguity here. Which port
does the filer use to reach the 192.168.5 network? It
seems to have picked one port (fortunately on the 'active'
switch) which I see listed in the routing table. The other
port is not listed in any routes at all.

But I was wondering if I could add explicit host routes
to the IP addresses on the standby switch. I tried this
and it didn't work (ping failed). The destination IP
that I tried to ping was on another filer head. It occurred
to me that I probably need to set the corresponding route
on the other head, so I did that. The routes are listed in
the route table, but ping still does not work. I checked
the ARP table and neither head has the MAC address of the
other in its ARP table. So I suppose that despite the host
route, the ARP broadcast is going out the wrong interface.

I think we need separate subnets here but I'm just the
storage guy and did not configure the hosts that are plugged
into these two switches.

Things are working, all traffic is on the active switch. But
I sure don't see how things could ever fail over to the
standby switch without manual intervention.

Steve Losen scl [at] virginia phone: 434-924-0640

University of Virginia ITC Unix Support

dleeds at edmunds

Apr 16, 2009, 1:58 PM

Post #4 of 4 (1612 views)
Permalink

RE: Network routing question [In reply to]

yes, i forgot to include the VSS/SMLT extensions (we use cisco switches).

http://en.wikipedia.org/wiki/SMLT

although rereading the original it sounds like they just want redundant links versus increased
bandwidth and high availability at the same time.

the standard vif with a vif favor will accomplish this and is quite simple. vif favor will also return
the connection to the primary switch when it's connection is restored rather than remain on the standby switch.

________________________________________
From: Robert Blayzor [rblayzor.bulk [at] inoc]
Sent: Thursday, April 16, 2009 1:47 PM
To: Leeds, Daniel
Cc: Stephen C. Losen; toasters [at] mathworks
Subject: Re: Network routing question

On Apr 16, 2009, at 4:20 PM, Leeds, Daniel wrote:
> if this is a redundant setup why would you not configure your
> switches with 802.3ad link aggregation and configure a
> vif on your filer with those two interfaces?

I believe for 802.3ad both connections would have to go to the same
physical switch.

--
Robert Blayzor, BOFH
INOC, LLC
rblayzor [at] inoc
http://www.inoc.net/~rblayzor/

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact Gossamer Threads