☠️ There is a better version of this module ready to be merged in the kernel (to appear in 4.18). A backport is available.

Based on Google's Maglev algorithm, this scheduler builds a lookup table in a way disruption is minimized when a change occurs. This helps in case of active/active setup without synchronization. Like for classic source hashing, this lookup table is used to assign connections to a real server.

Both source address and port are used to compute the hash (unlike sh where this is optional).

Weights are correctly handled. Unlike sh, servers with a weight of 0 are considered as absent. Also, unlike sh, when a server becomes unavailable due to a threshold, no fallback is possible: doing so would seriously impair the the usefulness of using a consistent hash.

The value of 65537 for the hash table size is currently not modifiable at compile-time. This is the value suggested in the Maglev paper. Another possible value is 257 (for small tests) and 655373 (for very large setups).

This is an out-of-tree module. Just type make and you should get an ip_vs_csh.ko file. You can use insmod to load it. If your kernel source are in a non-standard place, use make KDIR=....

There is no option for this scheduler. You can use it by its name:

ipvsadm -A -t 203.0.113.15:80 -s csh
ipvsadm -a -t 203.0.113.15:80 -r 10.234.79.11:80 -m
ipvsadm -a -t 203.0.113.15:80 -r 10.234.79.12:80 -m
ipvsadm -a -t 203.0.113.15:80 -r 10.234.79.13:80 -m -w 2