A5800 link aggragation

[aojea]

I want to configure a link aggregation with LACP between a CEntos Server and a HP 5800 with Version 5.20, Release 1206.

I think there is a problem with the load sharing method in the switch beacause I have connectivity with the server when I have only on link enabled. When I've got the 2 links enabled I can't reach the server with a ping, but in the server I can sniff traffic from the switch.Furthermore, when I configure "link-aggregation load-sharing mode source-ip destination-ip" I can reach the server with 7 pings, after the 7 ping I can't reach again.

I paste below the configuration of the switch:

interface Bridge-Aggregation2
 description CEntOS_Server
#

interface GigabitEthernet1/0/19
 port link-mode bridge
 port link-aggregation group 2
#
interface GigabitEthernet1/0/20
 port link-mode bridge
 port link-aggregation group 2

[Fred_Mancen_1]

Remove all the Interface Bridge settings, disconnect the cables from the ports. Then try to enter this template:

interface bridge 2

interface GigabitEthernet1/0/19
 port link-aggregation group 2

interface GigabitEthernet1/0/20
 port link-aggregation group 2

interface bridge 2

 description CEntOS_Server

 port link-mode bridge

Re-connect the cables and watch if the bridge-interface is up. Use the display link-aggregation verbose to check the aggregation group status. Make a double check in the server settings, some servers uses the LACP to bind interfaces. If this is necessary you will need to enable LACP in the bridge-interface too, using the command link-aggregation mode dynamic.

A few days ago I faced a similar problem; when I tried to set up the LAG in this way the aggregated link started to work. This occurs because the Bridge-aggregation interface acts as the main interface, reflecting all the settings in the selected ports of its link-aggregation group.

If you enter any command straight in the ports, not in the virtual bridge-aggregation port, the communication goes down. This is a particularity of v.5 software, that in this features acts as the Etherchannel (interface port-channel) in Cisco switches.

HTH.

[aojea]

I have done all steps but I have the same problem.

When the LAG is established it only works with load-sharing mode ="source-ip destination-ip" and I can reach only some IPs.

When I have only one link working in the LAG (shut down the other interface in the server), I can reach all IPs in the network.

[Fred_Mancen_1 · Accepted solution]

This is really weird. I already did these settings with LAG in the v.3 and v.5 and didn't have problems. Make a double check in the server settings; which binding method are you using in the server setup? Teaming or something like that? Did you tried to setup the bridge-aggregation without the load-sharing mode?

In my expertise with this feature, all the server settings were always responsible by manage the load-share itselves, the switch just manage the L2 connections, nothing else. Are you using multicast in the server?

Regards

[aojea]

Solved, I need to configure "link-aggregation mode dynamic"

[aojea]

I have solved using "link-aggregation mode dynamid"

Thanks

[paulgear]

I know you've solved the problem, but i thought i'd mention one point about this: you don't actually NEED switch-based link aggregation with Linux servers.  With the Linux bonding driver many of the options are available without configuring anything on the switch.  See http://www.howtoforge.com/nic-bonding-on-debian-lenny for more information (the link is specific to Debian, but the same options are available on CentOS as well).

[aojea]

I know that you say about bonding, but the linux server was only used to test without network disruption.

I have a Procurve 5406zl with 3 gb links to the same switch, and I want to do a 3 gb LAG between both switches.

I attach you a simple schematic.

Thanks for your help