A node of an LSP may inform the ingress node of the LSP, for example via RSVP signaling, about its temporary unavailability for a certain time. In response, the ingress node can stop using any affected LSP(s) and divert the traffic to other LSPs. This provides a faster mechanism to signal traffic shift then traditional IGP overload which causes considerable churn into the network as all the nodes need to compute the SPF. It is sufficient for ingress node to be aware of this node maintenance and it can use information to divert the traffic to other equal cost multipath (ECMP) LSP(s), or other available LSP(s). If no alternative LSP path exists when the ingress node receives such a message, a new LSP can be built during this time and traffic diverted smoothly (e.g., in a make-before-break manner) before the node goes offline for maintenance. Since only the ingress node is responsible to push the traffic to the LSP, there is no need to tear down the LSP for such node maintenance (especially when they are for a short duration). This can be used with a controller responsible for the LSP as well.