Configuring EtherChannels and Switch Troubleshooting Computer Networking Configuring EtherChannels and Switch Troubleshooting

EtherChannels STP is a great solution to avoid a broadcast storm in our network, but it effectively removes bandwidth we could use If we have two gigabit connections connecting two switches together, using STP I will be limited to only using one port and blocking the other EtherChannel technology allows us to bundle these connections together so that they are seen as one 2 gigabit per second connection If we do this, they will even be seen by STP as one connection and not be blocked This is a Cisco proprietary technology, and two protocols exist for it: PAgp (Port Aggregation Protocol) and LACP (Link Aggregation Control Protocol

PAgP VS LACP PAgP is used between two Cisco switches. It will create EtherChannels as well as prevents loops, limits packet loss due to misconfigured channels, and aid in network reliability. Each port will operate in two modes - Auto - the port is in a passive negotiating state and will form an EtherChannel if the port receives PAgP packets. The port will not initiate the negotiation Desirable - the port is placed in a negotiating state to form an EtherChannel by sending PAgP packets. A channel is formed with another port group in either the auto or desirable mode LACP is used between a Cisco switch and a non-Cisco switch that follows the IEEE 802.3ad standard. Ports can be in one of two modes as well - Passive - the port is put into a passive negotiating state and forms an EtherChannel if the port receives LACP packets. The port does not initiate the negotiation. This is equivalent to auto Active - the port is placed in a negotiating state to form an EtherChannel by sending LACP packets. A channel is formed with another port group in either the active or passive mode. This is equivalent to desirable

EtherChannel Commands The following commands will control our EtherChannel configuration at the port level #channel-protocol [lacp or pagp] - Selects the EtherChannel protocol on the interface #channel-group[1-8] mode [auto, desirable, active or passive] - Selects the PAgP or LACP mode on the interface #no channel-group[1-8] - disables the EtherChannel on the device We can also use show etherchannel at the switch level to get details about the etherchannels on the switch

EtherChannel troubleshooting Check that all ports are enabled Make sure both ports are using the same protocol, and that one side is either active or desirable Verify that both ports have the same speed and duplex. Not that LACP requires full duplex Check that the ports have the correct and matching channel numbers Verify that all ports are part of the same VLAN Check that none of the ports are blocked by STP before configuring the EtherChannel Make sure you are connecting the right number of ports. LACP EtherChannels can use up to 16 ethernet ports

Other Troubleshooting for Switches - Collisions Collisions can only occur if there are multiple devices sharing transmission media Each switch port is going to represent its own collision domain (meaning that we can’t have network collisions from outside this connection) This means that so long as the port is connected to an end point device like a workstation or server, or another switch we should never see a collision This means that there are only two possibilities if we are getting collisions We have a hardware failure causing transmissions when it shouldn’t We have a hub connected to a switch port. Since a hub broadcasts every frame it receives instead of selectively forwarding them our collision domain now includes every port on the hub and every device it is connected to This means that in the case of a collision, we should either isolate the damaged hardware and replace it or change out a hub with a switch

Duplexing Duplexing is whether we can receive and transmit at the same time on the same network medium. Full duplex means we can, half duplex means that we are limited to transmitting or receiving. This will affect our bandwidth When full duplex is used, the devices assume there can be no collisions since only two devices are connected. Therefore, not only is half duplex slower, but it will also cause the devices to check for collisions, slowing down the connection further By default most devices are going to be set to automatically configure their duplex mode when they are connected to another device

Mismatched Duplex Settings If one of the devices is in half duplex mode, and the other device is set to auto, the auto device will not get a response to it’s duplexing negotiation query and will default to half duplex, since it can’t be sure it is only connected to one device The problem is that it will also do this if one device is set to FULL duplex, which will create a duplex mismatch The half duplex device will attempt to detect collisions, and the full duplex device will not, and continuously send information. This will drastically hurt bandwidth To prevent this, keep all devices on auto if at all possible, but if you must manually set a duplex mode, set it for both sides of the connection

Link Speed Settings Most devices will try to automatically detect the appropriate link speed for their connection The connection will drop down to the lowest port setting of the two devices. For example, if I connect a NIC capable of 1 gigabit per second transmission to a 100 megabit switch port the connection will run at 100 megabit per second If a link is running at a speed slower than you expect, first check that the connection is not manually configured to run at a lower speed. If the ports are auto-configured, try setting the speed on both to manually be at the high speed you expect If the speed is set correctly and the link is still running slow, it is likely a hardware problem

Broadcast Storms Broadcast storms caused by network loops are very dangerous as they can eat up all of the bandwidth on a connection and take down a network We can prevent this from happening by using the Spanning Tree Protocol, or STP Many high-end switches will also include special software that will detect broadcast storms, and drop some broadcast traffic to allow normal communications to occur

Misconfigured VLAN assignments Remember that each separate VLAN on a switch will act as its own network, and that VLANs are assigned by switch port This means that if we were to change the port a device is connected to, it is possible that we would be changing its VLAN, and no longer be able to connect to its proper network In fact, it is commonly considered best-practice to set unconnected ports to their own separate VLAN so that people connecting rogue devices won’t be connected to other devices If a device is unable to see the devices it should be connected to, check to make sure its switch port is on the correct VLAN

Frame Errors Frame errors are when a switch sees a frame it cannot process, and it drops it In general we will see one of three frame issues Oversized - this usually means that the NIC that generated this frame is gibbering, sending out junk data that can’t be processed, and not sending a frame with a proper end signal Undersized - this usually means there was a collision and the frame lost part of its data CRC - The CRC check does not match the data. This could occur for a number of reasons, but definitely indicates the message has been corrupted No matter what the error, the frame will be dropped. If this keeps happening it will drastically slow down our connection since the dropped frames will need to be resubmitted The switch will record these dropped frames, so if there is a slow connection we can examine the switch and see if this is the issue