Commentary

weave

@@ -285,14 +285,22 @@ attach() {
     arp_update $CONTAINER_IFNAME $1 "ip netns exec $NETNS"
 }
 
-# Configure the ARP cache parameters for the given interface
+# Configure the ARP cache parameters for the given interface.  This
+# makes containers react more quickly to a change in the MAC address
+# associated with an IP address.
 configure_arp_cache() {
     $2 sh -c "echo 5 >/proc/sys/net/ipv4/neigh/$1/base_reachable_time &&
               echo 2 >/proc/sys/net/ipv4/neigh/$1/delay_first_probe_time &&
               echo 1 >/proc/sys/net/ipv4/neigh/$1/ucast_solicit"
 }
 
-# Send out an unsolicated ARP response to flush ARP cache entries
+# Send out an unsolicated ARP response to update ARP cache entries
+# accross the weave network.  We do this in addition to
+# configure_arp_cache because a) with those ARP cache settings it
+# still takes a few seconds to correct a stale ARP mapping, and b)
+# there is a kernel bug that means that the base_reachable_time
+# setting is not promptly obeyed
+# (<https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=4bf6980dd0328530783fd657c776e3719b421d30>>).
 arp_update() {
     # It's not the end of the world if this doesn't run - we configure
     # ARP caches so that stale entries will be noticed quickly.