0

I have a device that is connected to a switch and is then connected to a customer controller. I have been asked to make the connection to the controller redundant because they are very far apart from one another. In the case of one connection failing there would be another to take over that is routed through a different area. I do not want the increased throughput, only the fail over protection.

I only have access to my device and the managed switch. I have no access to the controller and the customer will not have access to the switch. I have read about LACP and LAG but I'm a programmer not a network admin so I'm not quite sure how this process works. The switch to buy is being based off this decision so I have no hardware to work with. It seems to me that the configuration would have to be done on the controller OS and then also I would have to configure the two ports on my switch to reflect that configuration. Is this the proper way to achieve this setup?

EDIT: To clarify what I'm asking. If its something that can be setup on just the switch, then we would probably do it. If it requires messing with the switch and the controller NICs, then we would probably tell the customer to buy their own switch.

ewwhite
  • 197,159
  • 92
  • 443
  • 809
rawbus
  • 3
  • 2
  • No offense @rawbus, i know this comment is not going to answer your question but seriously hire someone, you don't need a top-notch engineer for this nor expensive at all :) – Cold T Jun 18 '14 at 20:59
  • What model of managed switch? What Operating System is the "controller" running? Is the controller a Windows/Linux server? Does it have multiple network interfaces? If so what is the model of the Ethernet controllers? – myron-semack Jun 18 '14 at 21:07
  • More information. Is this industrial equipment? How far is the switch from the controller? – ewwhite Jun 18 '14 at 21:10
  • @msemack the switch is to be determined. – rawbus Jun 18 '14 at 21:10
  • @ewwhite yes it is industrial and you could say across the plant floor. I don't know say a 100 yards. – rawbus Jun 18 '14 at 21:11
  • Your "controller" needs two Ethernet ports, and they will both need to be configured properly. Either that or you have managed switches at both ends and just do a redundant uplink between the two switches. – myron-semack Jun 18 '14 at 21:16
  • I'm assuming their controller would or he wouldn't be asking me about link layer level redundancy. – rawbus Jun 18 '14 at 21:19
  • And I'm guessing the device only has one Ethernet port? – myron-semack Jun 18 '14 at 21:20
  • Correct that is why the switch is involved. – rawbus Jun 18 '14 at 21:21

3 Answers3

0

If you really don't need the added throughput of LACP you just need the configure the controller to use a master/slave (the terminology differs depending on brand/driver) or Adaptive Load Balancing. Neither of these configurations require any configuration on the switch side,but provide L1 redundancy. You can also provide L2 redundancy here by having the lines run into two separate switches.

Zypher
  • 37,405
  • 5
  • 53
  • 95
0

I have industrial and production equipment like this in my environments. Unless you have existing cabling or other hardware issues, I wouldn't go out of my way to provide connection resiliency. We don't in our configurations (except for that location with the rodent-chewing-through cables problem).

  • Do you happen to know what type of controller this is?
  • What Operating System does it run?
  • What are the manufacturer's best-practices or recommendations regarding managing the network interfaces?
  • Can you elaborate on the application. That will help us understand the tolerance for failure/downtime and possible workarounds.

Think about your potential issues and steps to remediate them...

  • Switch failure: Replace switch. Keep cold-spare.
  • Cable failure: Replace cable. Run additional cabling as a precaution.
  • Controller interface failure: Use another available interface(?)

Am I missing any potential issues here?

ewwhite
  • 197,159
  • 92
  • 443
  • 809
  • I'm not going to get any information on the controller it could be anything. But it is safe to assume it has two ports and supports some type of redundancy or he wouldn't be asking me if the switch does. The application is controlling a motor. If the link goes down the motor control goes down. The main worry is cable failure because of the distance. We are not worried about switch or controller interface. – rawbus Jun 18 '14 at 22:44
  • Initially the redundancy was going to be done in software by treating the two connections from the controller as separate interfaces. But the customer inquired about handling the redundancy in the switch. – rawbus Jun 18 '14 at 22:49
0

Ok, based on the comments, here is what I would do. It is crude, but will work and is really simple to configure.

Use a Cisco switch with the switchport backup interface configuration command. Designate one port on the switch to be the primary uplink, and another one to be the secondary uplink. http://www.cisco.com/c/en/us/td/docs/switches/lan/catalyst3750/software/release/12-2_55_se/commmand/reference/3750cr/cli3.html#wp3269214

As long as the primary interface has a link, the backup interface will be kept offline (no link LED). If the primary interface goes down (cable is cut), it will be placed in an offline state, and the backup interface will be brought online.

Now, the controller will need to be connected to the the primary and backup interfaces of the switch, and the network interfaces on the controller will need to be configured to use the other port in the case of a link down event. If this is not practical, put another switch in front of the controller so it only has to deal with one link. (No special configuration of the controller-side switch will be needed. Could be an unmanaged switch.)

             |      |----------------------------|      |
Controller---|Switch|                            |Switch|---Device
             |      |----------------------------|      |
myron-semack
  • 2,593
  • 19
  • 17