Troubleshooting 802.3ad Dynamic LAG Connections
Options
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Report Inappropriate Content
‎11-05-2013 05:02 PM
Article ID: 5340
Products
DFE
Matrix C1
SecureStack C3
SecureStack C2
SecureStack B3
SecureStack B2
SecureStack A2
Matrix E1
Standards
802.3ad
Symptoms
LAG does not establish
LAG disconnects
Communication errors over the LAG
Solution
Common Layer 1 and 2 problems can cause Dynamic LAGs to function inappropriately. The 802.3ad protocol needs a stable Layer 1 and 2 environment to establish and maintain connectivity. The following bullet items may be useful for isolating issues related to Dynamic LAG establishment and maintenance.
Spanning Tree
A LAG is a redundant connection until the 802.3ad protocol establishes two or more links as a single logical LAG. It is important for Spanning Tree to be enabled on the underlying ports to avoid a physical loop occurring while the LAG is not operational.
Configuration Problems vs Intermittent Issues
Generally, a configuration issue is a show stopper, not allowing normal connectivity to occur. Most likely causes of configuration issues is manual keying of the LAGs, and VLAN misconfigurations.
Intermittent issues, except if the underlying ports are misconfigured, are generally not due to misconfiguration. LAGS only work with similarly configured full duplex ports, so generally configuration is not an issue at the port level. Intermittent issues may be caused by poor cabling (which will act worse as activity increases), Spanning Tree instability in the network, and other environmental conditions which may keep the protocol or underlying ports from maintaining connectivity.
Short and Long LAG timers
The default timers for the lag are "long". The protocol transmits maintenance packets every 30 seconds. This means that the protocol will wait up to 3 X 30 seconds for a failure to occur if there is no link loss. "Short" timers send the protocol every second, and shorten the failover time to three seconds in the above circumstances. Using short timers is appropriate in environments where a link loss locally, such as a carrier loss from a media converter or a cable modem (channelizing connection), is not likely. In these cases, the protocol or secondary connections are not passing protocol or data, but because of a lack of link loss, the underlying ports are not removed from the LAG, and communication errors occur.
In these environments, lowering the lag timers to "short" on both ends of the LAG will help quicken carrier failover times.
N-Series DFE:
set port lacp port fe.1.1-2 aadminstate lacptimeout
Matrix C1:
set port lacp port fe.1.1-2 aadminstate lacptimeout
(use clear instead of set to remove)
SecureStacks:
set port lacp port fe.1.1-2 aadminstate lacptimeout
(use clear instead of set to remove)
Matrix E1:
set port lacp aadminstate fe.0.1-2 short-timeout
For ports that are participating in the LAG; in addition to setting the actor admin state to short timeout (facilitates ports dropping out of the LAG quicker), set the partner adminstate to short timeout (enables port to rejoin the LAG much quicker when LACP PDUs return), as well.
The difference in time it takes ports to rejoin the LAG is significant: Up to 30 seconds if the partner admin state is set to long timeout, vs approximately 1 second if set to short timeout, on both ends of the LAG.
N-Series DFE:
set port lacp port fe.2.1-2 padminstate lacptimeout
Matrix C1:
not settable
SecureStacks:
set port lacp port fe.2.1-2 padminstate lacptimeout
(use clear instead of set to remove)
Matrix E1:
not settable
Diagnosing LAG issues
What follows are some example command sets to identify problems at layer 1 and 2.
N-Series DFE (lag.0.x), Matrix C1 (lag.1.x), and SecureStacks (lag.0.x):show spantree stats
show port status fe.2.1-48;lag.0.2 [underlying ports should be dormant]
show spantree debug
show spantree debug active
show spantree debug port fe.2.1-48;lag.0.2
show port lacp port fe.2.1-48 counters
show port lacp port fe.2.1-48 status detail
show lacp [will display all lags]
show port counters fe.2.1-48
show port counters lag.0.1
show rmon stats fe.2.1-48
show vlan
show vlan static
Matrix E1:show spantree stats
show port status fe.0.1-2 [underlying ports should be dormant]
show port counters fe.0.1-2
show rmon stats fe.0.1-2
show port lacp detail fe.0.1-2
show port lacp counters fe.0.1-2
show vlan
show vlan static
Contact Enterasys Networks Technical Services as necessary, for further assistance.
Products
DFE
Matrix C1
SecureStack C3
SecureStack C2
SecureStack B3
SecureStack B2
SecureStack A2
Matrix E1
Standards
802.3ad
Symptoms
LAG does not establish
LAG disconnects
Communication errors over the LAG
Solution
Common Layer 1 and 2 problems can cause Dynamic LAGs to function inappropriately. The 802.3ad protocol needs a stable Layer 1 and 2 environment to establish and maintain connectivity. The following bullet items may be useful for isolating issues related to Dynamic LAG establishment and maintenance.
Spanning Tree
A LAG is a redundant connection until the 802.3ad protocol establishes two or more links as a single logical LAG. It is important for Spanning Tree to be enabled on the underlying ports to avoid a physical loop occurring while the LAG is not operational.
Configuration Problems vs Intermittent Issues
Generally, a configuration issue is a show stopper, not allowing normal connectivity to occur. Most likely causes of configuration issues is manual keying of the LAGs, and VLAN misconfigurations.
Intermittent issues, except if the underlying ports are misconfigured, are generally not due to misconfiguration. LAGS only work with similarly configured full duplex ports, so generally configuration is not an issue at the port level. Intermittent issues may be caused by poor cabling (which will act worse as activity increases), Spanning Tree instability in the network, and other environmental conditions which may keep the protocol or underlying ports from maintaining connectivity.
Short and Long LAG timers
The default timers for the lag are "long". The protocol transmits maintenance packets every 30 seconds. This means that the protocol will wait up to 3 X 30 seconds for a failure to occur if there is no link loss. "Short" timers send the protocol every second, and shorten the failover time to three seconds in the above circumstances. Using short timers is appropriate in environments where a link loss locally, such as a carrier loss from a media converter or a cable modem (channelizing connection), is not likely. In these cases, the protocol or secondary connections are not passing protocol or data, but because of a lack of link loss, the underlying ports are not removed from the LAG, and communication errors occur.
In these environments, lowering the lag timers to "short" on both ends of the LAG will help quicken carrier failover times.
N-Series DFE:
set port lacp port fe.1.1-2 aadminstate lacptimeout
Matrix C1:
set port lacp port fe.1.1-2 aadminstate lacptimeout
(use clear instead of set to remove)
SecureStacks:
set port lacp port fe.1.1-2 aadminstate lacptimeout
(use clear instead of set to remove)
Matrix E1:
set port lacp aadminstate fe.0.1-2 short-timeout
For ports that are participating in the LAG; in addition to setting the actor admin state to short timeout (facilitates ports dropping out of the LAG quicker), set the partner adminstate to short timeout (enables port to rejoin the LAG much quicker when LACP PDUs return), as well.
The difference in time it takes ports to rejoin the LAG is significant: Up to 30 seconds if the partner admin state is set to long timeout, vs approximately 1 second if set to short timeout, on both ends of the LAG.
N-Series DFE:
set port lacp port fe.2.1-2 padminstate lacptimeout
Matrix C1:
not settable
SecureStacks:
set port lacp port fe.2.1-2 padminstate lacptimeout
(use clear instead of set to remove)
Matrix E1:
not settable
Diagnosing LAG issues
What follows are some example command sets to identify problems at layer 1 and 2.
N-Series DFE (lag.0.x), Matrix C1 (lag.1.x), and SecureStacks (lag.0.x):show spantree stats
show port status fe.2.1-48;lag.0.2 [underlying ports should be dormant]
show spantree debug
show spantree debug active
show spantree debug port fe.2.1-48;lag.0.2
show port lacp port fe.2.1-48 counters
show port lacp port fe.2.1-48 status detail
show lacp [will display all lags]
show port counters fe.2.1-48
show port counters lag.0.1
show rmon stats fe.2.1-48
show vlan
show vlan static
Matrix E1:show spantree stats
show port status fe.0.1-2 [underlying ports should be dormant]
show port counters fe.0.1-2
show rmon stats fe.0.1-2
show port lacp detail fe.0.1-2
show port lacp counters fe.0.1-2
show vlan
show vlan static
Contact Enterasys Networks Technical Services as necessary, for further assistance.
0 REPLIES 0
