The input block (max 1) allows you to configure how BGP updates are processed when received from a peer. A key feature is the ability to filter updates before they are parsed and stored in memory, which significantly reduces memory usage.
Memory-Efficient Filtering
Use these parameters to drop unwanted updates early in the process:
accept_comunities / filter_communities: Filter by BGP communities.accept_ext_communities / filter_ext_communities: Filter by extended communities.accept_large_comunities / filter_large_communities: Filter by large communities.accept_nlri / filter_nlri: Filter by specific NLRIs (requires session restart).accept_unknown / filter_unknown: Filter by unknown attributes.
Note: Using accept_* parameters reduces memory usage by discarding updates before they reach the routing table. Using standard filter chains will still consume memory as prefixes are marked as not active, filtered in the routing table.
Input Performance and Limits
affinity: Configures multi-core processing for input. Options include alone (best for many cores/peers), afi, instance, vrf, remote-as, main, or input (runs output in the same process as input).limit_process_routes_ipv4 / limit_process_routes_ipv6: Limits the number of received routes to prevent memory exhaustion. If the limit is reached, a BGP session clear command is required to reset the flag.