Showing results for 
Search instead for 
Did you mean: 

IRF Resilience

Occasional Visitor

IRF Resilience

Hi everyone,


A quick question on resilience of IRF stacks on E48/A5500 switches.  I have a stack of 4 x E4800G switches all up and running fine in one of my datacentres, but whilst undertaking maintenance on the power supplies have had a real question mark placed over stack resilience in the event of unit failures.


For info:

Unit 1   Slave    Priority 1

Unit 2   Slave    Priority 2

Unit 3   Master  Priority 3

Unit 4   Slave    Priority 1


My plan was to power down and move each unit individually over to a temporary power supply whilst the maintenance work was taking place, which would hopefully mean that my servers (with multiple connections across the stack) would stay up at all times.  However, i found that rebooting unit 2 (slave) also caused unit 1 to reboot.  When unit 3 (master) was rebooted, it caused unit 2 and 1 to reboot again!  When the units all returned to full operation, unit 1 is now the Master device - the lowest priority device!?


I thought units only rebooted when first being added to the virtual device and topology, etc was being replicated... surely an already formed IRF stack should not exhibit this behaviour?


I raise this point as it doesnt strike me as particularly resilient if in a power blip or failure situation, repowering or replacing a unit (be it master or slave) means other units will be disrupted as well.  Has anyone else found this to be the case?