From: Faulconer, Steven M (steven.m.faulconer@lmco.com)
Date: Tue Dec 13 2005 - 12:53:30 EST
Hello everyone,
I've been tasked with investigating trunking as a possible replacement
for IPMP (we need performance in addition to failover). I SEEM to have
gotten everything setup, but I'm having one persistent error and some
oddities in performance (which my be a lack of understanding or due to
the previously mentioned error). First, my configuration:
Host A is a 280R with a 4-port GigaSwift card running Solaris 8
Host B is a V890 with a 4-port GigaSwift card running Solaris 10
Switch is a Cisco 3550
On A and B, I have a very similar nettr -setup script:
nettr -setup 0 device=ce members=0,1,2,3 policy=2
Except B has the key as 2 since it already has several ce interfaces
configured.
HostA 'nettr -conf':
------------------------------------------------------------------------
------------
Key: 0; Policy: 2;
Aggr MAC address: 0:3:ba:da:19:37
Name Original-Mac-Addr Speed Duplex Link Status
---- ----------------- ----- ------ ---- ------
ce0 0:3:ba:da:19:37 100 full up enb
ce1 0:3:ba:da:19:38 100 full up enb
ce2 0:3:ba:da:19:39 100 full up enb
ce3 0:3:ba:da:19:3a 100 full up enb
------------------------------------------------------------------------
------------
HostB 'nettr -conf':
------------------------------------------------------------------------
------------
Key: 2; Policy: 2;
Aggr MAC address: 0:3:ba:b3:4:30
Name Original-Mac-Addr Speed Duplex Link Status
---- ----------------- ----- ------ ---- ------
ce2 0:3:ba:b3:4:30 100 full up enb
ce3 0:3:ba:b3:4:31 100 full up enb
ce4 0:3:ba:b3:4:32 100 full up enb
ce5 0:3:ba:b3:4:33 100 full up enb
------------------------------------------------------------------------
------------
Now, on the switch, I ran through this setup to create the
EtherChannel:
------------------------------------------------------------------------
------------
configure terminal
interface range fa0/# - #
switchport trunk encapsulation dot1q
switchport mode trunk
channel-group # mode auto
------------------------------------------------------------------------
------------
Now HostA is on ports 1-4 and HostB is on ports 11-14. Also, HostA is
on channel-group 1, where HostB is on channel-group 2.
Now, the error I'm getting is that another MAC address is taking our
IP. I've changed the BootProm variable 'local-mac-address?' to be true
on both systems. Things work (sort of) but I'm not sure if this error
might be causing my other issue, that being performance.
Running 'nettr -stats <key> interval=5' and then using rcp to copy a
1GB file between the systems shows the in and out packets being split
between the 4 interfaces (as one would expect in the round-robin
policy). However, I never break more than about 80 Mbit running one
copy. Running three copies simultaneously, I hit about 103 Mbit max.
This is confusing to me as I thought I should see over 300 Mbit. I know
I won't get 400 Mbit total (overhead and what not) but I expected
something more than this.
Any advice would be greatly appreciated. I imagine I'm having a mental
disconnect somewhere. Please let me know if you need additional
information.
Thanks,
Steven
(I posted this to the usenet (comp.sys.sun.admin) as well, sorry if you
see it in both places)
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 23:37:38 EDT