Comware

Hi Nicolas,

IRF is not a typical old-fashioned management clustering technology where every member has its own IP and you have just one common WebUI listening on a common for all switches IP address (added in addition to a unique IP per slot). IRF is a stacking technology where all physical members are treated as one chassis-based switch with two redundant MPUs where:

Master switch's control plane becomes something like a Master MPU in a chassis
All other switches' control plane becomes a sort of a single virtualised Standby MPU in that chassis
All physical ports of each members becomes like virtual line cards in a chassis - ports on the first member are 1/0/* (like a line card in Slot 1 of a chassis), ports on the second member are 2/0/* (like a line card in Slot 2 of a chassis) etc.

That is why you cannot have an IP address to access each particular member - it doesn't make any sense, it's like "I want to ssh to my linecard in Slot 3", it's just not possible. Each IRF member outsources its brain 😉 to the current Master, the only work that performs this "brain" (control plane) is syncing with the Master, listening to configuration and management commands coming from the Master and reporting various local events to the Master. Even if you get a console cable, connect it to a console port of, let's say member 3 which is a Standby at the moment, you will get the command prompt from the Master (member 1, I suppose in your case). The IP 10.100.3.10 will always belong to the current Master, no matter which IRF member is that.

Regarding MAD. If you will get a split brain and suddenly let's say member 2 will become a new Master, but member 1 will keep its Master role, then suddenly BOTH IPs 10.100.3.11 AND 10.100.3.13 will become active, the BFD session between them will be established and it will be a clear sign of the split-brain situation. Then the member 1 will signal over the MAD session to member 3 to disable its ports and accept its destiny to obey and be Standby 🙂)) When a stack is healthy, from all that MAD IPs only one IP is active, therefore BFD session is DOWN, therefore it is a signal that everything is OK and there is only one Master aboard 🙂

So your expectations are wrong, you just need more time to accommodate yourself to all these new concepts.

Hi Nicolas,

IRF is not a typical old-fashioned management clustering technology where every member has its own IP and you have just one common WebUI listening on a common for all switches IP address (added in addition to a unique IP per slot). IRF is a stacking technology where all physical members are treated as one chassis-based switch with two redundant MPUs where:

Master switch's control plane becomes something like a Master MPU in a chassis
All other switches' control plane becomes a sort of a single virtualised Standby MPU in that chassis
All physical ports of each members becomes like virtual line cards in a chassis - ports on the first member are 1/0/* (like a line card in Slot 1 of a chassis), ports on the second member are 2/0/* (like a line card in Slot 2 of a chassis) etc.

That is why you cannot have an IP address to access each particular member - it doesn't make any sense, it's like "I want to ssh to my linecard in Slot 3", it's just not possible. Each IRF member outsources its brain 😉 to the current Master, the only work that performs this "brain" (control plane) is syncing with the Master, listening to configuration and management commands coming from the Master and reporting various local events to the Master. Even if you get a console cable, connect it to a console port of, let's say member 3 which is a Standby at the moment, you will get the command prompt from the Master (member 1, I suppose in your case). The IP 10.100.3.10 will always belong to the current Master, no matter which IRF member is that.

Regarding MAD. If you will get a split brain and suddenly let's say member 2 will become a new Master, but member 1 will keep its Master role, then suddenly BOTH IPs 10.100.3.11 AND 10.100.3.13 will become active, the BFD session between them will be established and it will be a clear sign of the split-brain situation. Then the member 1 will signal over the MAD session to member 3 to disable its ports and accept its destiny to obey and be Standby 🙂)) When a stack is healthy, from all that MAD IPs only one IP is active, therefore BFD session is DOWN, therefore it is a signal that everything is OK and there is only one Master aboard 🙂

So your expectations are wrong, you just need more time to accommodate yourself to all these new concepts.