I've setup a vritual chassis which includes four members. (Two linecard role, 1 master re, and 1 backup re). I did failover testing for this virtual chassis and found some problems. Below is my testing procedure. (Virtual Chassis is in non-preprovisioning mode)
1. Member 1 has a 255 mastership-priority. (Master RE)
Member 2 has a 254 mastership-priority. (Backup RE)
Member 0 has an 128 mastership-priority. (Linecard)
Member 3 has an 128 mastership-priority. (Linecard)
2. I shutdown three members as the following sequence.
Member 1, Member 0, Member3
3. I expected Member 2 will be the master re after shutdown member3. But the result was not what I expected. I found that member 2 became linecard instead of Master RE.
What's wrong with it?
In addition, I found that when master re was shutdown, all port other switch member will become inactive until new Master RE was elected. Can I shorten the inactive period?
It could be how you have you VCP ports configured, do you have a complete loop where every switch is connected to another. If not you might have a break in which case there might not be a path back to your Backup RE. Also, to help with failover convergence, try enabling Graceful Routing Engine Switchover.
Your stack shows a loop which is good. Since your switch 2 is backup, it should become master after switch 1 goes down, then since 0 and 3 have same priority then 3 might becomme backup due to lowest MAC. How quickly are you shutting down these switches, are you waiting 5 seconds, 1 minute or 5-10 minutes to give time for switch to go down or are you just removing poower from device to testing.
The solution was to use the "set virtual-chassis no-split-detection" statement but this should normaly be used when there are only two devices, but since you are shutting down all the 3 other devices in your stack of 4, I believe that this statement could solve your problem !