03-08-2012 09:22 AM
We are fairly sure there is something wrong with our existing Master switch. I would like to issues the command "request chassis routing-engine master switch" but we are also in a live, high production enviroment. How long will the network drop for? This is two EX4200 in a VC acting as our network core. We also run 24x7 so downtime is not easy to find.
Note I will be chaning the current master priority from 255 to 128 so that it matches the second switch in the VC shortly as well.
Any feedback appriciated.
Solved! Go to Solution.
03-08-2012 10:25 AM
depends on a couple of factors --
1) are you using the built-in VCP cables, or extended via fiber?
2) is the VC layer-2 only, or performing layer 3 functions as well?
3) what version of code are you running? newer versions have added NSR and NSB which will greatly affect the downtime (should be next to none with these running)
if its layer 2 only then downtime should be very brief -- if STP is in use it will have to recalculate though (unless NSB is being used). if its also functioning as a router, and NSR is not being used, then routing adjacencies will bounce, so convergence really depends on what protocols you are using.
03-08-2012 10:27 AM
also I forgot to add -- make sure you have GRES configured regardless, otherwise downtime will be way longer as the line-card switches have to reconnect to the new master, relearn MACs, etc.
03-08-2012 01:01 PM
in theory I believe if you have graceful restart enabled and all of the surrounding routers support it and have it enabled, traffic loss should be very small. I do not have much experience with graceful restart, however, so can't really comment. and I think enabling requires an adjacency reset (may be protocol specific, not sure)/
at any rate, I would say to be safe expect however much time it would take if you were to just reset all of your routing adjacencies on the VC, as that is what will happen when the routing engines flip, since RPD will have to restart on the new master.