Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations strongm on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!
Do I need two different name for two ethernet card?
- Thread starter redbub
- Start date
First card smit mininimum configuration second card further configuration
and set up the card......Different networks for each card...you can add a name to the second card in the host file if you wish......One hostname....
--------------------------------------------------------
Implications of Multiple Network Interfaces on the Same
Network
Sometimes network managers feel the need to provide greater availability and
performance by adding a second network adapter to a particular machine. For
example, they may want to have two token-ring adapters attached to the
same physical network. Which it is possible to have more than one interface
on the same network, in general this is not recommended for two reasons:
1) Having two interfaces on the same network is a violation of TCP/IP
architecture.
In TCP/IP architecture, a host machine with two network adapters is defined
as an IP router. Different network adapters must be attached to different
physical networks. In the case of token-ring, TCP/IP addresses multiple rings
bridged together as a single logical ring(as if it were a single physical ring)
2) Having two interfaces on the same network can cause broadcast storms.
Whenever an IP host sees traffic for a network whose IP address is different
from its own network, it generates an Internet Control Message Protocol
(ICMP) packet announcing this conflict. Since every host on the network sees
the traffic that is miss addressed, every host generates ICMP packets. If the
amount of miss addressed traffic is significant, the ICMP traffic an grow to the
point that network performance degrades. It is possible to avoid the broadcast
storms created when multiple interfaces are connected to the same network.
However, doing so is still a violation of TCP/IP architecture. You must as you
know assign distinct IP addresses and names to tr0 and tr1. TCP/IP
architecture requires that each interface have a unique IP address and name;
otherwise, unpredictable behavior will result. For instance, you might give
interface tr0 IP address 10.10.10.1 and the name laurel.foo.bar.com and
interface tr1 the IP address 10.10.10.2 and the name hardy.foo.bar.com...but
it is still a violation. Attempts to set up the above 10.10.10.1 and 10.10.10.2 in
4.3.2 and above usually provides errors and caused us not to be able to
configure it. Perhaps IBM has written in the configuration rules.
Another Option in 4.3.3 is:
Cisco Etherchannel Support
How does etherchannel work? Etherchannel would operate at layer 2 below
the protocol stack. The EtherChannel would have one MAC address and for IP
it would have one IP address. For outgoing traffic, the packets are sent out
over one of the adapters of the channel. Which adapter to use is decided by
the lower bits of destination MAC address of the outgoing packet for non-IP
traffic. For IP traffic, the lower bits of destinatioin IP address is used to decide
the adapter. When an adapter of an Etherchannel is down, traffic would be
re-routed through one of the other adapters and this would be transparent to
the upper layers. The Etherchannel can be configured with up to 4 Ethernet
adapters. It provides a scaleble server to switch bandwidth. It allows the
combination of multiple ethernet adapters together to form a larger pipe. The
etherchannel allows a server to switch connection throughput between 40 and
800 Mbps.
The etherChannel can be configured with up to 4 ether adapters (ent0, ent2,
ent3, ent5 ) By default all the adapters of the etherchannel would be assigned
the MAC address of the first adapter that the user selects. Software comes
as a part of AIX 4.3.3 and is devices.common.IBM.ethernet.rte. It is set up in
smit under smitty etherchannel. You will then need to select two to four
adapters from the list to form your channel.
One has to remember that trunking solutions such as EtherChannel need
routers which are capable of talking the same language. It doesn't matter if
you have AIX 4.3.3 running EtherChannel or Solaris 2.7 running Trunking 1.x
or HPUX 11.x with Port Aggregation software. If you don't have the correct
model of Cisco router with the appropriate software on it, you can't use the
trunking software on your OS.
and set up the card......Different networks for each card...you can add a name to the second card in the host file if you wish......One hostname....
--------------------------------------------------------
Implications of Multiple Network Interfaces on the Same
Network
Sometimes network managers feel the need to provide greater availability and
performance by adding a second network adapter to a particular machine. For
example, they may want to have two token-ring adapters attached to the
same physical network. Which it is possible to have more than one interface
on the same network, in general this is not recommended for two reasons:
1) Having two interfaces on the same network is a violation of TCP/IP
architecture.
In TCP/IP architecture, a host machine with two network adapters is defined
as an IP router. Different network adapters must be attached to different
physical networks. In the case of token-ring, TCP/IP addresses multiple rings
bridged together as a single logical ring(as if it were a single physical ring)
2) Having two interfaces on the same network can cause broadcast storms.
Whenever an IP host sees traffic for a network whose IP address is different
from its own network, it generates an Internet Control Message Protocol
(ICMP) packet announcing this conflict. Since every host on the network sees
the traffic that is miss addressed, every host generates ICMP packets. If the
amount of miss addressed traffic is significant, the ICMP traffic an grow to the
point that network performance degrades. It is possible to avoid the broadcast
storms created when multiple interfaces are connected to the same network.
However, doing so is still a violation of TCP/IP architecture. You must as you
know assign distinct IP addresses and names to tr0 and tr1. TCP/IP
architecture requires that each interface have a unique IP address and name;
otherwise, unpredictable behavior will result. For instance, you might give
interface tr0 IP address 10.10.10.1 and the name laurel.foo.bar.com and
interface tr1 the IP address 10.10.10.2 and the name hardy.foo.bar.com...but
it is still a violation. Attempts to set up the above 10.10.10.1 and 10.10.10.2 in
4.3.2 and above usually provides errors and caused us not to be able to
configure it. Perhaps IBM has written in the configuration rules.
Another Option in 4.3.3 is:
Cisco Etherchannel Support
How does etherchannel work? Etherchannel would operate at layer 2 below
the protocol stack. The EtherChannel would have one MAC address and for IP
it would have one IP address. For outgoing traffic, the packets are sent out
over one of the adapters of the channel. Which adapter to use is decided by
the lower bits of destination MAC address of the outgoing packet for non-IP
traffic. For IP traffic, the lower bits of destinatioin IP address is used to decide
the adapter. When an adapter of an Etherchannel is down, traffic would be
re-routed through one of the other adapters and this would be transparent to
the upper layers. The Etherchannel can be configured with up to 4 Ethernet
adapters. It provides a scaleble server to switch bandwidth. It allows the
combination of multiple ethernet adapters together to form a larger pipe. The
etherchannel allows a server to switch connection throughput between 40 and
800 Mbps.
The etherChannel can be configured with up to 4 ether adapters (ent0, ent2,
ent3, ent5 ) By default all the adapters of the etherchannel would be assigned
the MAC address of the first adapter that the user selects. Software comes
as a part of AIX 4.3.3 and is devices.common.IBM.ethernet.rte. It is set up in
smit under smitty etherchannel. You will then need to select two to four
adapters from the list to form your channel.
One has to remember that trunking solutions such as EtherChannel need
routers which are capable of talking the same language. It doesn't matter if
you have AIX 4.3.3 running EtherChannel or Solaris 2.7 running Trunking 1.x
or HPUX 11.x with Port Aggregation software. If you don't have the correct
model of Cisco router with the appropriate software on it, you can't use the
trunking software on your OS.
ElgisRamon
MIS
AIXQueen, where did you find it a host with two interfaces in the same TCP/IP network is violating the standards?
You can have your AIX like this:
[tt]
en0 = 200.201.202.203 = machinename.domain.com
en1 = 200.201.200.204
[tt]
These settings can be set up using smit.
Then go to your DNS server and set two A[/a] pointers for the name machinename.domain.com, then whenever a client machine asks for the name it can get any or both names and try the connection.
Many internet servers are set like this, try getting the IP addresses for a big site like or
The BIND name server has settings to round robin the answers to clients (the named daemon included in AIX is BIND).
I hope it works...
Unix was made by and for smart people.
You can have your AIX like this:
[tt]
en0 = 200.201.202.203 = machinename.domain.com
en1 = 200.201.200.204
[tt]
These settings can be set up using smit.
Then go to your DNS server and set two A[/a] pointers for the name machinename.domain.com, then whenever a client machine asks for the name it can get any or both names and try the connection.
Many internet servers are set like this, try getting the IP addresses for a big site like or
The BIND name server has settings to round robin the answers to clients (the named daemon included in AIX is BIND).
I hope it works...
Unix was made by and for smart people.
It is from the IBM site............
--------------------------------------------------
Implications of Multiple Network Interfaces on the Same Network
Sometimes network managers feel the need to provide greater availability and
performance by adding a second network adapter to a particular machine. For
example, they may want to have two token-ring adapters attached to the same
physical network. While it is possible to have more than one interface on the same
network, in general this is not recommended for two reasons:
1.Having two interfaces on the same network is a violation of TCP/IP architecture.
In TCP/IP architecture, a host machine with two network adapters is defined as
an IP router. Different network adapters must be attached to different physical
networks. In the case of token-ring, TCP/IP addresses multiple rings bridged
together as a single logical ring (as if it were a single physical ring).
2.Having two interfaces on the same network can cause broadcast storms.
Whenever an IP host sees traffic for a network whose IP address is different
from its own network, it generates an Internet Control Message Protocol (ICMP)
packet announcing this conflict. Since every host on the network sees the traffic
that is misaddressed, every host generates ICMP packets. If the amount of
misaddressed traffic is significant, the ICMP traffic can grow to the point that
network performance degrades.
It is possible to avoid the broadcast storms created when multiple interfaces
are connected to the same network. However, doing so is still a violation of TCP/IP
architecture. The solution is to give the different interfaces different IP addresses
on the same network. For example, you might have two token-ring adapters on the
same network named tr0 and tr1. You must assign distinct IP addresses and names
to tr0 and tr1. (TCP/IP architecture requires that each interface have a unique IP
address and name; otherwise, unpredictable behavior will result.) For instance,
you might give interface tr0 the IP address 10.10.10.1 and the name
laurel.foo.bar.com, and interface tr1 the IP address 10.10.10.2 and
the name hardy.foo.bar.com.
------------------------------------------
He did not DNS or HACMP.............or subnetting........and since he is a beginner
I am assuming that he is doing /etc/hosts..........In the past with out a different
subnet, the smit commands failed if you tried to configure two interfaces with
the similar IPs on the same network...perhaps it lets you do it now in 4.3.3....but
you can see IBM's note above.
--------------------------------------------------
Implications of Multiple Network Interfaces on the Same Network
Sometimes network managers feel the need to provide greater availability and
performance by adding a second network adapter to a particular machine. For
example, they may want to have two token-ring adapters attached to the same
physical network. While it is possible to have more than one interface on the same
network, in general this is not recommended for two reasons:
1.Having two interfaces on the same network is a violation of TCP/IP architecture.
In TCP/IP architecture, a host machine with two network adapters is defined as
an IP router. Different network adapters must be attached to different physical
networks. In the case of token-ring, TCP/IP addresses multiple rings bridged
together as a single logical ring (as if it were a single physical ring).
2.Having two interfaces on the same network can cause broadcast storms.
Whenever an IP host sees traffic for a network whose IP address is different
from its own network, it generates an Internet Control Message Protocol (ICMP)
packet announcing this conflict. Since every host on the network sees the traffic
that is misaddressed, every host generates ICMP packets. If the amount of
misaddressed traffic is significant, the ICMP traffic can grow to the point that
network performance degrades.
It is possible to avoid the broadcast storms created when multiple interfaces
are connected to the same network. However, doing so is still a violation of TCP/IP
architecture. The solution is to give the different interfaces different IP addresses
on the same network. For example, you might have two token-ring adapters on the
same network named tr0 and tr1. You must assign distinct IP addresses and names
to tr0 and tr1. (TCP/IP architecture requires that each interface have a unique IP
address and name; otherwise, unpredictable behavior will result.) For instance,
you might give interface tr0 the IP address 10.10.10.1 and the name
laurel.foo.bar.com, and interface tr1 the IP address 10.10.10.2 and
the name hardy.foo.bar.com.
------------------------------------------
He did not DNS or HACMP.............or subnetting........and since he is a beginner
I am assuming that he is doing /etc/hosts..........In the past with out a different
subnet, the smit commands failed if you tried to configure two interfaces with
the similar IPs on the same network...perhaps it lets you do it now in 4.3.3....but
you can see IBM's note above.
It is from the IBM site............
--------------------------------------------------
Implications of Multiple Network Interfaces on the Same Network
Sometimes network managers feel the need to provide greater availability and
performance by adding a second network adapter to a particular machine. For
example, they may want to have two token-ring adapters attached to the same
physical network. While it is possible to have more than one interface on the same
network, in general this is not recommended for two reasons:
1.Having two interfaces on the same network is a violation of TCP/IP architecture.
In TCP/IP architecture, a host machine with two network adapters is defined as
an IP router. Different network adapters must be attached to different physical
networks. In the case of token-ring, TCP/IP addresses multiple rings bridged
together as a single logical ring (as if it were a single physical ring).
2.Having two interfaces on the same network can cause broadcast storms.
Whenever an IP host sees traffic for a network whose IP address is different
from its own network, it generates an Internet Control Message Protocol (ICMP)
packet announcing this conflict. Since every host on the network sees the traffic
that is misaddressed, every host generates ICMP packets. If the amount of
misaddressed traffic is significant, the ICMP traffic can grow to the point that
network performance degrades.
It is possible to avoid the broadcast storms created when multiple interfaces
are connected to the same network. However, doing so is still a violation of TCP/IP
architecture. The solution is to give the different interfaces different IP addresses
on the same network. For example, you might have two token-ring adapters on the
same network named tr0 and tr1. You must assign distinct IP addresses and names
to tr0 and tr1. (TCP/IP architecture requires that each interface have a unique IP
address and name; otherwise, unpredictable behavior will result.) For instance,
you might give interface tr0 the IP address 10.10.10.1 and the name
laurel.foo.bar.com, and interface tr1 the IP address 10.10.10.2 and
the name hardy.foo.bar.com.
------------------------------------------
He did not say DNS or HACMP.............or subnetting........and since he is a beginner
I am assuming that he is doing /etc/hosts..........In the past with out a different
subnet, the smit commands failed if you tried to configure two interfaces with
the similar IPs on the same network...perhaps it lets you do it now in 4.3.3....but
you can see IBM's note above.
--------------------------------------------------
Implications of Multiple Network Interfaces on the Same Network
Sometimes network managers feel the need to provide greater availability and
performance by adding a second network adapter to a particular machine. For
example, they may want to have two token-ring adapters attached to the same
physical network. While it is possible to have more than one interface on the same
network, in general this is not recommended for two reasons:
1.Having two interfaces on the same network is a violation of TCP/IP architecture.
In TCP/IP architecture, a host machine with two network adapters is defined as
an IP router. Different network adapters must be attached to different physical
networks. In the case of token-ring, TCP/IP addresses multiple rings bridged
together as a single logical ring (as if it were a single physical ring).
2.Having two interfaces on the same network can cause broadcast storms.
Whenever an IP host sees traffic for a network whose IP address is different
from its own network, it generates an Internet Control Message Protocol (ICMP)
packet announcing this conflict. Since every host on the network sees the traffic
that is misaddressed, every host generates ICMP packets. If the amount of
misaddressed traffic is significant, the ICMP traffic can grow to the point that
network performance degrades.
It is possible to avoid the broadcast storms created when multiple interfaces
are connected to the same network. However, doing so is still a violation of TCP/IP
architecture. The solution is to give the different interfaces different IP addresses
on the same network. For example, you might have two token-ring adapters on the
same network named tr0 and tr1. You must assign distinct IP addresses and names
to tr0 and tr1. (TCP/IP architecture requires that each interface have a unique IP
address and name; otherwise, unpredictable behavior will result.) For instance,
you might give interface tr0 the IP address 10.10.10.1 and the name
laurel.foo.bar.com, and interface tr1 the IP address 10.10.10.2 and
the name hardy.foo.bar.com.
------------------------------------------
He did not say DNS or HACMP.............or subnetting........and since he is a beginner
I am assuming that he is doing /etc/hosts..........In the past with out a different
subnet, the smit commands failed if you tried to configure two interfaces with
the similar IPs on the same network...perhaps it lets you do it now in 4.3.3....but
you can see IBM's note above.
redbub,
Place the second adapter in a different logical subnet via smitty chinet.
i.e. retain the same subnet mask but change the last network octet.
Then vi /etc/hosts and place a different hostname in the IP address definition.
Cheers
PSD
IBM Certified Specialist - AIX V4.3 Systems Support
IBM Certified Specialist - AIX V4 HACMP
Place the second adapter in a different logical subnet via smitty chinet.
i.e. retain the same subnet mask but change the last network octet.
Then vi /etc/hosts and place a different hostname in the IP address definition.
Cheers
PSD
IBM Certified Specialist - AIX V4.3 Systems Support
IBM Certified Specialist - AIX V4 HACMP
I personally like the etherchannel idea, as it is the way I am currently trying to set up my pSeries 660. The idea about having the name server dishing out the two IP addresses alternately sounds intriguing, but since I have no control over the DNS server, I'm going to stick to Etherchannel.
My question is, has anyone else tried etherchannel? I have two ethernet NIC's, one on-board, and one as an adapter. I want to use etherchannel more for thoughput than redundancy, but redundancy is important also. AIX etherchannel set-up has 3 modes, standard, round-robin, and ifbackup.
Ifbackup mode will use one adapter until it dies, and then switch over to the over. This mode is out. Standard mode takes the outbound IP address, does a hashing algorhythm, and picks a enet card to use for that IP address, and only uses that card for that outbound address. Round-robin alternately uses each card in the pair for maximum bandwidth utilization. Round-robin is the mode that I want to use.
I went and set up the etherchannel pair as a round-robin pair, got the network guys to set the switch as etherchannel, and voila, it works great. Here's my dilemma. I set up the pseries to ping the default gateway, pull one of the 2 nic cables, and only half of the packets get transmittted. I plug the nic cable back in, pull the other, and again, only half get transmitted.
I then tried standard mode for the etherchannel pair. It does only use one card for each IP. Same problem though, when you pull the cable for the nic it's using, it doesn't fail to the other.
Any ideas?
My question is, has anyone else tried etherchannel? I have two ethernet NIC's, one on-board, and one as an adapter. I want to use etherchannel more for thoughput than redundancy, but redundancy is important also. AIX etherchannel set-up has 3 modes, standard, round-robin, and ifbackup.
Ifbackup mode will use one adapter until it dies, and then switch over to the over. This mode is out. Standard mode takes the outbound IP address, does a hashing algorhythm, and picks a enet card to use for that IP address, and only uses that card for that outbound address. Round-robin alternately uses each card in the pair for maximum bandwidth utilization. Round-robin is the mode that I want to use.
I went and set up the etherchannel pair as a round-robin pair, got the network guys to set the switch as etherchannel, and voila, it works great. Here's my dilemma. I set up the pseries to ping the default gateway, pull one of the 2 nic cables, and only half of the packets get transmittted. I plug the nic cable back in, pull the other, and again, only half get transmitted.
I then tried standard mode for the etherchannel pair. It does only use one card for each IP. Same problem though, when you pull the cable for the nic it's using, it doesn't fail to the other.
Any ideas?
- Status