Techniques are described for managing failover in redundant network devices. In particular, each device in a set of redundant network devices includes redundant processing modules. Each module provides a separate operating environment for a set of network services. Each network device includes a control unit that receives configuration information that specifies individual weight values for each of the modules and a threshold value for the network device. The control unit detects failures of the modules and transfers primary responsibility for performing the network service to a second network device when the threshold value is exceeded by a weighted sum of the weight values for the failed modules.