Re: Gigabit Ether Channel
2003-01-11 20:57:45
Our adventure in Etherchannel seems to have been successful.
smitty etherchannel;
Add an Etherchannel
Type or select values in entry fields.
Press Enter AFTER making all desired changes.
[Entry Fields]
Etherchannel Adapters ent0 ent3 ent4 ent5 +
Enable ALTERNATE ETHERCHANNEL address no +
ALTERNATE ETHERCHANNEL address [] +
Mode standard +
Enable GIGABIT ETHERNET JUMBO frames no +
Internet Address to Ping []
Number of Retries [] #
Retry Timeout (sec) [] #
F1=Help F2=Refresh F3=Cancel F4=List
F5=Reset F6=Command F10=Edit F8=Image
F9=Shell F10=Exit Enter=Do
The output was
ent6 Available
Then it was as simple as assigning a TCP/IP address to en6.
Testing using FTP from several hosts with Gigabit Ethernet Adapters, we
did:
Ftp to this new adapter's address
ftp> put "|dd if=/dev/zero bs=1024k count=30000 " /dev/null
We then watched the entstat -d ent6 output to see the distribution of the
packets.
Thanks for everyone's help.
----- Forwarded by Jeff G Kloek/IT/IPAPER on 01/11/2003 07:26 PM -----
Jeff G Kloek
To: "ADSM: Dist Stor
Manager" <ADSM-L AT VM.MARIST DOT EDU>
01/11/2003 01:34 cc:
PM Subject: Re: Gigabit Ether
Channel(Document link: Jeff G Kloek)
Direct from IBM:
A little further research reveals that when you set up etherchannel via the
smitty etherchannel menus you are by default setting up cisco base port
aggregation.
We'll try this later tonight and I'll report on the results.
Thanks for all your help.
"Chris Murphy"
<cmurphy AT IDL DOT STAT To: ADSM-L AT
VM.MARIST DOT EDU
E.ID.US> cc:
Sent by: "ADSM: Subject: Re: Gigabit Ether
Channel
Dist Stor
Manager"
<[email protected]
.EDU>
01/10/2003 01:25
PM
Please respond to
"ADSM: Dist Stor
Manager"
>>Please hazard a guess on this one: If the adapter doesn't and the channel
doesn't form, does that mean the ip address won't be pingable at all?
Thanks
again for your help - it's much appreciated. Jeff
In my experience, if the channel does not form, the switch has no way of
knowing these 2 (or more) ports go to the same node. Therefore, it will
treat them as two (or more) separate nodes since it will see two different
MAC addresses: one on each port. The IP address assigned to each NIC
should
then be ping-able (asuming all routing/VLANs and such working properly) as
it would as if you were not trying to channel and just had a node with 2
NICs. The only way I can think of this NOT happening, is if:
(a) mis-config of NIC(s) on node e.g. no IP assigned, administratively
downed
(b) the switch is set to FORCE a channel (done with "CHANNEL-GROUP x MODE
ON" or some variation of depending on switch model...). In this case, the
switch will be expecting a channel to form, if one does not, connectivity
may not be established. This can be used to form a channel if the NIC(s)
do
NOT suppport PAgP, or have weak support of, but are still capable of
EtherChannel since no PAgP frames are sent in the "ON" mode.
We had case (b) happen on some Intel NICs we had once. They supported
EtherChannel, but for some reason PAgP did not work. When we set the
channel to "ON" (forced) and it worked. This might be the case with ours.
Hope that helps some!
Chris
|
|
|