Status now sent with gzip - libbpf bumped to v1.5.0

Status is now requested gzipped; large configurations lead to huge JSON documents being requested every couple of seconds, meaning many Mb/s of data being consumed when idly watching the status page. This should reduce the bandwidth consumed to a couple of percent of that which was used before.

This might prevent some nasty surprises on mobile phone roaming bills (in future :-/ ...)

Healthchecks and flags

HTTP/S probes will use the VIP as the host in the URL for checks if none is specified in check definition.
New cue version uses cached HTTP connections by default (avoids high numbers of TIME_WAIT connections)
Flags updated to match those in the stayinalived (IPVS/NAT) variant.

Flag/config changes and internals

v0.2.17

capricious flag/config file changes

Migrated common code from main to library

v0.2.16

code moved to library portion from main

Bumped cue library version to fix eBGP empty AS_PATH bug

v0.2.15

bumper cur library version to fix eBGP empty AS_PATH bug

Logging hot reload

Logging code re-written. Now able to reload logging config on the fly - previously a full restart was required.

Command channel

Added a UNIX domain socket option (-C) to allow sending commands to the running daemon - used to reattach XDP code to interfaces initially

Support for changing the maximum number of flows

v0.2.11

support for changing max number of flows

Loopback BGP option

Enables the use of a routing daemon (eg. BIRD) on the load balancer server, advertising VIPs to it on 127.0.0.1

See docs/bgp.md for more details.

Prometheus stats issue fixed

Prometheus stats were showing the total stats instead of individual values for each backend