Traffic Route
This policy allows us to configure routing rules for L4 traffic running in our Mesh. This policy provides support for weighted routing and can be used to implement versioning across our services as well as deployment strategies like blue/green and canary.
TrafficRoute
must select the data plane proxies to route the connection between them.
Kuma also supports locality aware load balancing.
Default TrafficRoute
The control plane creates a default TrafficRoute
every time the new Mesh
is created. The default TrafficRoute
enables the traffic between all the services in the mesh.
apiVersion: kuma.io/v1alpha1
kind: TrafficRoute
mesh: default
metadata:
name: route-all-default
spec:
sources:
- match:
kuma.io/service: '*'
destinations:
- match:
kuma.io/service: '*'
conf:
loadBalancer:
roundRobin: {}
split:
- weight: 100
destination:
kuma.io/service: '*'
type: TrafficRoute
name: route-all-default
mesh: default
sources:
- match:
kuma.io/service: '*'
destinations:
- match:
kuma.io/service: '*'
conf:
loadBalancer:
roundRobin: {}
split:
- weight: 100
destination:
kuma.io/service: '*'
Usage
By default when a service makes a request to another service, Kuma will round robin the request across every data plane proxy belogning to the destination service. It is possible to change this behavior by using this policy, for example:
apiVersion: kuma.io/v1alpha1
kind: TrafficRoute
mesh: default
metadata:
name: route-example
spec:
sources:
- match:
kuma.io/service: backend_default_svc_80
destinations:
- match:
kuma.io/service: redis_default_svc_6379
conf:
split:
- weight: 90
destination:
kuma.io/service: redis_default_svc_6379
version: '1.0'
- weight: 10
destination:
kuma.io/service: redis_default_svc_6379
version: '2.0'
We will apply the configuration with kubectl apply -f [..]
.
type: TrafficRoute
name: route-example
mesh: default
sources:
- match:
kuma.io/service: backend
destinations:
- match:
kuma.io/service: redis
conf:
split:
- weight: 90
destination:
kuma.io/service: redis
version: '1.0'
- weight: 10
destination:
kuma.io/service: redis
version: '2.0'
We will apply the configuration with kumactl apply -f [..]
or via the HTTP API.
In this example the TrafficRoute
policy assigns a positive weight of 90
to the version 1.0
of the redis
service and a positive weight of 10
to the version 2.0
of the redis
service.
Note that routing can be applied not just on the automatically provisioned service
Kuma tag, but on any other tag that we may want to add to our data plane proxies (like version
in the example above).
Kuma utilizes positive weights in the TrafficRoute
policy and not percentages, therefore Kuma does not check if the total adds up to 100. If we want to stop sending traffic to a destination service we change the weight
for that service to 0.
Load balancer types
There are different load balancing algorithms that can be used to determine how traffic is routed to the destinations. By default TrafficRoute
uses the roundRobin
load balancer, but more options are available:
roundRobin
is a simple algorithm in which each available upstream host is selected in round robin order.Example:
loadBalancer:
roundRobin: {}
leastRequest
uses different algorithms depending on whether the hosts have the same or different weights. It has a single configuration fieldchoiceCount
, which denotes the number of random healthy hosts from which the host with the fewer active requests will be chosen.Example:
loadBalancer:
leastRequest:
choiceCount: 8
ringHash
implements consistent hashing to the upstream hosts. It has the following fields:hashFunction
the hash function used to hash the hosts onto the ketama ring. Can beXX_HASH
orMURMUR_HASH_2
.minRingSize
minimum hash ring size.maxRingSize
maximum hash ring size.
Example:
loadBalancer:
ringHash:
hashFunction: "MURMUR_HASH_2"
minRingSize: 64
maxRingSize: 1024
random
selects a random available host.Example:
loadBalancer:
random: {}
maglev
implements consistent hashing to upstream hostsExample:
loadBalancer:
maglev: {}