REP convergence failure

[Mark Wardle]

I have 2-off IE-3000-4TC access switches and a stacked 3750X in a REP ring with the following configuration;

3750X

rep admin vlan 31
int gi1/0/1
switchport mode trunk
switchport trunk native vlan 31
rep segment 1 edge primary

int gi2/0/1
switchport mode trunk
switchport trunk native vlan 31
rep segment 1 edge preferred

IE-3000-4TC

rep admin vlan 31
int range gi1/1-2
switchport mode trunk
switchport trunk native vlan 31
rep segment 1

IE-3000-4TC

rep admin vlan 31
int range gi1/1-2
switchport mode trunk
switchport trunk native vlan 31
rep segment 1

show rep top

3750X gi1/0/1 Pri Open
IE-3000-4TC gi1/1 Open
IE-3000-4TC gi1/2 Open
IE-3000-4TC gi1/1 Open
IE-3000-4TC gi1/2 Open
3750X gi2/0/1 Sec Alt

When I remove either link connecting the IE-3000-4TC access switches to the edge ports of the 3750X, REP converges successfully. However if I remove the trunk link between the 2 IE-3000-4TC access switches, the Alt port does not open, REP does not converge and I lose the second IE-3000-4TC access switch from the ring.

Any advice would be greatly appreciated.

[Furose M · Accepted solution]

Hi Mark,

the config looks good and pretty basic. are you able to consistently repro the issue?

what is the output of "show rep topology" from 3750x and IE switch after disconnecting the link?

will it be possible to test with default admin vlan and default native vlan?

[Mark Wardle]

Yes, same issue is reproduced each time I remove a link not connected to an edge port.  I have since configured 3 IE-3000 switches in a REP ring using default native vlan and same issue occurs.

Output with edge link removed

show rep topology
IE3000    gi1/1               Fail
IE3000    gi1/2               Open
IE3000    gi1/1               Open
IE3000    gi1/2               Open
3750X     gi2/0/1      Sec    Open

Output with trunk link between access switches removed

show rep topology
IE3000    gi1/1               Fail
IE3000    gi1/2               Open
3750X     gi2/0/1     Sec     Open

The following messages are generated when that trunk link is removed (none of these appear when an edge link is removed);

*Mar 1 01:22:14.900: %REP-4-LINKSTATUS: GigabitEthernet1/2 (segment 1) is non-o
perational due to port status change
*Mar 1 01:22:15.890: %LINEPROTO-5-UPDOWN: Line protocol on Interface GigabitEth
ernet1/2, changed state to down
*Mar 1 01:22:16.905: %LINK-3-UPDOWN: Interface GigabitEthernet1/2, changed stat
e to down

[Furose M · Accepted solution]

where was this output collected?

show rep topology
IE3000    gi1/1               Fail
IE3000    gi1/2               Open
3750X     gi2/0/1     Sec     Open

i guess on 3750X. i can see the port is in Open state. where did you lose access to IE from?

you mean, even after the link is set to "Open", it was not passing traffic?

[Mark Wardle]

Yes, that output was from the 3750X.  Please find the output taken from every switch in the second lab I configured (3 IE-3000 switches in REP ring) below; taken after trunk link removed.

S2#show rep top
REP Segment 1
Warning: REP detects a segment failure, topology may be incomplete
BridgeName PortName Edge Role
---------------- ---------- ---- ----
S2 Gi1/2 Fail
S2 Gi1/1 Open
S1 Gi1/1 Sec Open

S1#show rep top
REP Segment 1
Warning: REP detects a segment failure, topology may be incomplete
BridgeName PortName Edge Role
---------------- ---------- ---- ----
S1 Gi1/1 Sec Open
S2 Gi1/1 Open
S2 Gi1/2 Fail

S3#show rep top
REP Segment 1
Warning: REP detects a segment failure, topology may be incomplete
BridgeName PortName Edge Role
---------------- ---------- ---- ----
S3 Gi1/2 Fail
S3 Gi1/1 Open
S1 Gi1/2 Sec Open

Now I've done that, I'm a little confused as to whether it is working or not, as when all switch outputs are taken in to consideration, all 3 switches appear in the topology.  Is this what I should expect?

[Mark Wardle]

Configured an access vlan this morning and added some end devices.  Ping tested those devices whilst removing links and got zero loss.  After all this, REP was working the entire time.  I guess I was thrown by the missing switch when only looking at the show rep topology output from one switch in the ring.

I expected to see a consistent rep topology across all switches, which obviously is not the case.  It makes sense once you put the outputs from all switches together, as you requested, above.  Thanks very much for your assistance.  It is much appreciated. 

[Furose M]

YEs. this is what we should expect when a link failure occurs.