From UBNT:
New features:
Offloading – Add CLI commands to disable flow-table flushing in offloading engine when routing table changes: set system offload ipv4 disable-flow-flushing-upon-fib-changes set system offload ipv6 disable-flow-flushing-upon-fib-changes
Discussed here
Prior to 1.10.3 firmware flow-table in offloading engine was always flushed when route was updated in linux routing table. Flow flushing ensured that offloading engine got routing updates instantly but it wasted a lot of CPU time and decreased performance if routing table was constantly updated for (instance in Full BGP, big OSPF or flapping PPPoE interface scenarios)
In 1.10.3 firmware by default disable-flow-flushing-upon-fib-changes is not set which means that flow table in offloading engine is always flushed upon routing table changes same way as it used to be in previous firmware.
If you have Full-BGP table or large OSPF network they you are advised to set disable-flow-flushing-upon-fib-changes this will ensure less CPU-load and increase max throughput.
Important note for multi-WAN environments – if nexthop interface of default-gateway changes and disable-flow-flushing-upon-fib-changes is set then it will take up to flow-lifetime seconds before all existing offloaded flows switch to new nexthop interface (up to 12 seconds by default).
Offloading – Add CLI command to modify flow-lifetime in offloading engine (expressed in seconds):
set system offload flow-lifetime 24Prior to 1.10.3 firmware flow-lifetime parameter was hardcoded and was not synchronized between different ER platforms: 12 seconds on ER-Lite/ER-Poe, 6 seconds on ER/ER-pro/ER-4/ER-6 and 3 seconds on ER-Infinity.
In 1.10.3 firmware default value of flow-lifetime is set to 12 seconds for all ER platforms and now it can be modified. By modifying flow-lifetime parameter you control how much traffic skips from offloading engine into linux network stack.
If you increase flow-lifetime then:
a) Offloaded IP flows will expire less frequently and less packets will be forwarded to linux
b) CPU load will decrease and max throughput will increase
c) if disable-flow-flushing-upon-fib-changes parameter is set then it will take more time for offloading engine to detect changes in routing table
If you decrease flow-lifetime then:
a) Offloaded IP flows will expire more frequently and more packets will be forwarded to linux
b) CPU load will increase and max throughput will decrease
c) if disable-flow-flushing-upon-fib-changes parameter is set then it will take less time for offloading engine to detect changes in routing table
Offloading – add CLI command to show flows in offloading engine: show ubnt offload flows Offloading – add CLI command to show offloading engine statistics: show ubnt offload statistics
Enhancements and bug fixes:
LDP – fixed regression in 1.10.0 when LDP configuration failed. Discussed here LoadBalancing – fixed regression in 1.10.1 when LoadBalancing failed to recover if WAN interface lost&restored link in 3 second interval. Discussed here DHCP – fixed bug when DHCP server configuration failed to commit with networks other than /8, /16, and /24. Discussed here TrafficControl – fixed regression in 1.10.0 when “command not found” output was printed when running “show traffic-control …” commands. Discussed here
