Removal of odd numbered supervisor when in standby mode can cause traffic interupts

Bug CSCvo56267 

In-band path from one or more line cards to MDS 9700 supervisor is mis-programmed. The consequence is that control plane traffic sourced from this line card is not properly routed. This may lead to various symptoms that impact data plane, such as:

- FSPF hellos from peer switches are missed, leading to routes to remote domains being withdrawn
- Interfaces are unable to come up

Other types of irregular behavior may also be possible.

This issue applies only to odd-numbered supervisor slots on MDS 9700 platform. Supervisor slots on different switch models that are part of this platform are as follows:

MDS 9718: 9,10
MDS 9710: 5,6
MDS 9706: 3,4

This issue can be triggered only if, while lower-numbered slot supervisor is in ha-standby state, it is pulled out of the chassis without issuing the command `out-of-service module x`. It cannot be triggered if the lowered-numbered slot supervisor is active, or if the higher-numbered slot is pulled out of the chassis.

Although the issue is caused by the supervisor removal, traffic impact might not be triggered immediately, but only after a subsequent supervisor switchover, which may happen much later.

Workaround:
Always perform the out-of-service procedure before physically removing the supervisor from the chassis, as per MDS 9700 Hardware Installation Guide:
 

Where slot indicates the chassis slot number in which the standby supervisor module resides.

switch(config)# out-of-service module X

https://www.cisco.com/c/en/us/td/docs/switches/datacenter/mds9000/hw/9700/mds_9700_hig/replace.html#42939 


If you suspect that the traffic on your switch is already impacted by this issue reboot the switch, which should resolve the problem.

Further Problem Description:
Only physical removal of the supervisor can trigger this issue. Other types of logical removals such as supervisor reload (even if caused by a software crash) cannot cause it.