#1  
Old December 21st, 2007, 02:02 PM
jshatch jshatch is offline
Junior Member
NETGEAR Newbie
 
Join Date: Dec 2007
Posts: 6
jshatch is on a distinguished road
Default GS108T and link aggregation

Hello.

I'm trying to get link aggregation working properly between dual e1000 ports on a server running Fedora 8 and the GS108T switch.

I've set up link aggregation on ports 1 and 2 on the switch, and setup bond0 on the server with mode=4 (802.3ad).

I believe the configuration is correct on the server because I can get the link to work properly, but what I'm running into is that initially after a reboot the bond does not negotiate a link. Only after taking down the bond0 interface and the slave interfaces and bringing them up again (sometimes several times) will the link come up properly.

It almost seems like there is a timeout or something happening but I haven't tested that far. I've tried about every setting on the switch without any change in behavior.

If I set up the bond with mode=0 it comes up right away. But that's just round robin and I'd like to make it work with 802.3ad. Any tips on this?
Reply With Quote
  #2  
Old December 26th, 2007, 07:21 AM
chrwei chrwei is offline
Junior Member
NETGEAR Newbie
 
Join Date: Dec 2007
Posts: 8
chrwei is on a distinguished road
Default Re: GS108T and link aggregation

I'm having similar issues with a GS724T using Ubuntu. What I notice is that on the linux side I'll get this message in the syslog:
Quote:
Warning: No 802.3ad response from the link partner for any adapters in the bond
sometimes I ge the warning for just one of the slaves, in which case the link works, and sometimes for both in which case the link doesn't work. However, when I ping from the server to any other device on the network, a link will come up, but all traffic only comes from or goes to one of the slaves, no matter how many clients are connected.

In my case, on the switch under Trunking I have checked ID=2 and ports 9 and 11. I also upgraded the firmware to the lastest this last weekend with no improvement.
Reply With Quote
  #3  
Old December 26th, 2007, 07:29 AM
chrwei chrwei is offline
Junior Member
NETGEAR Newbie
 
Join Date: Dec 2007
Posts: 8
chrwei is on a distinguished road
Default Re: GS108T and link aggregation

Oh, also my server is a SuperMicro with embeded dual Intel 80003ES2LAN nics.
Reply With Quote
  #4  
Old December 26th, 2007, 08:08 AM
chrwei chrwei is offline
Junior Member
NETGEAR Newbie
 
Join Date: Dec 2007
Posts: 8
chrwei is on a distinguished road
Default Re: GS108T and link aggregation

per another post on this forum I disabled Fast Link, I already had STP disabled, and I no longer get the warning, but it is still not aggregating.

Using iperf on the server with the bonding and from 2 clients that do not have bonding. First test after disabling fastlink and reloading the binding driver I got 932 Mbps, which I'd call good. powered up the 2nd host and tested it, this time I got 227 Mbps, which is not good. verified all links were gigabit full duplux and they are.

here's where it gets weird, I start one client and get about 225Mbps, start the second and it gets about 225Mbps for about 3 seconds and then both jump to 450Mbps each, then if i stop one client the other takes the full 900+. I tried running a single client, either one, for 60 seconds on its own and it never goes above 230Mbps, but if I start the second client, even just for a couple seconds, it holds at about 930Mbps for the rest of the time.

I don't know if I'm seeing a linux issue or a netgear issue, but some searching doesn't reveal these types of problems with other brands of switches.
Reply With Quote
  #5  
Old December 29th, 2007, 07:44 PM
jshatch jshatch is offline
Junior Member
NETGEAR Newbie
 
Join Date: Dec 2007
Posts: 6
jshatch is on a distinguished road
Default Re: GS108T and link aggregation

I also see the following error a few times:

Code:
bonding: bond0: Warning: No 802.3ad response from the link partner for any adapters in the bond
I always get this error on both interfaces until I do the following:

1) wait for the machine to boot up
2) /etc/init.d/network restart
3) Abort the DHCP client with Ctrl-C
4) /etc/init.d/network restart

After the second restart the interface usually comes up and gets a DHCP address.

I haven't seen any difference with tweaking settings on the switch. I've tried with STP on and off, changed the lacp timeout from long to short, and even turned off flow control and the behavior remains the same.

Here's the lines I have in modprobe.conf regarding the bonding module:
Code:
alias bond0 bonding
options bond0 mode=4 miimon=100
Here's my ifcfg-bond0:
Code:
DEVICE=bond0
ONBOOT=yes
BOOTPROTO=dhcp
USERCTL=no
and here's the ifcfg-eth0 (eth1 is the same):
Code:
# Intel Corporation 82546GB Gigabit Ethernet Controller
DEVICE=eth0
BOOTPROTO=none
ONBOOT=yes
MASTER=bond0
SLAVE=yes
USERCTL=no
And finally, here is the output from /proc/net/bonding/bond0 after I get the link up:
Code:
Ethernet Channel Bonding Driver: v3.1.3 (June 13, 2007)

Bonding Mode: IEEE 802.3ad Dynamic link aggregation
Transmit Hash Policy: layer2 (0)
MII Status: up
MII Polling Interval (ms): 100
Up Delay (ms): 0
Down Delay (ms): 0

802.3ad info
LACP rate: slow
Active Aggregator Info:
        Aggregator ID: 3
        Number of ports: 2
        Actor Key: 17
        Partner Key: 18194
        Partner Mac Address: 00:1b:2f:2e:9b:3f

Slave Interface: eth0
MII Status: up
Link Failure Count: 0
Permanent HW addr: 00:30:48:2c:ac:26
Aggregator ID: 3

Slave Interface: eth1
MII Status: up
Link Failure Count: 0
Permanent HW addr: 00:30:48:2c:ac:27
Aggregator ID: 3
If anyone sees any weirdness that I'm missing please let me know.

Quote:
Originally Posted by chrwei
However, when I ping from the server to any other device on the network, a link will come up, but all traffic only comes from or goes to one of the slaves, no matter how many clients are connected.
Now that's kindof different for me. I see traffic on both slaves. At least, the counters increment on both ports on the switch as well as in ifconfig, though they are not incrementing evenly. I have not had the chance to test the performance of multiple machines hitting the server at the same time, though I was able to get ~870Mb/sec from a single machine using netpipe.

Quote:
Originally Posted by chrwei
Oh, also my server is a SuperMicro with embeded dual Intel 80003ES2LAN nics.
I also have a Supermicro, a X6DHE-G2 with dual Intel 82546GB ports. Not that I think that's the issue though...


Quote:
Originally Posted by chrwei
I don't know if I'm seeing a linux issue or a netgear issue, but some searching doesn't reveal these types of problems with other brands of switches.
Unfortunately I haven't been able to test with another switch. I might be able to get my hands on a Procurve and test with it. That would help point at a config issue vs. a switch issue. Since we are seeing this problem on 2 different switches from Netgear it might be an implementation issue. We'll see.
Reply With Quote
  #6  
Old December 30th, 2007, 10:04 PM
jshatch jshatch is offline
Junior Member
NETGEAR Newbie
 
Join Date: Dec 2007
Posts: 6
jshatch is on a distinguished road
Default Re: GS108T and link aggregation

Does anyone know what the configuration option "LACP System Setting" is, what the values are and why it defaults to 3842?
Reply With Quote
  #7  
Old January 4th, 2008, 06:32 AM
chrwei chrwei is offline
Junior Member
NETGEAR Newbie
 
Join Date: Dec 2007
Posts: 8
chrwei is on a distinguished road
Default Re: GS108T and link aggregation

I emailed support and they say my GS724T doesn't support LACP and so won't work with 802.3ad dynamic link aggregation. anyone know if the bonding driver supports manual trunking? so far my searches haven't yeilded anything possitive
Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off



All times are GMT -8. The time now is 09:10 PM.