This article describes the issue of 'request system reboot <all-chassis> ' on the SFC backup causing traffic to be disrupted, when the Re mastership is inconsistent.
In normal conditions, the RE mastership is consistent on SFC and LCC. the Junos upgrade can be implemented via the following procedure:
For multi-chassis TXP to operate without any traffic loss, it is absolutely necessary to keep the path, from SFC-Primary to LCC-Primary, up all the time. When SFC-RE0 and LCC-RE1 are the masters, there are two redundant paths between them:
SFC-RE0 <--igbe1--> SFC-CB1 <----> SFC-CIP1 <----> LCC-CB1 <--bcm0--> LCC-RE1
SFC-RE0 <--igbe0--> SFC-CB0 <----> SFC-CIP0 <----> LCC-CB0 <--em1--> LCC-RE1
Workaround: