Mixing nodes and pods
Table of contents
Scenario
A single scenario can have actions on both nodes and pods (don’t forget to configure a cloud driver).
For example, imagine that you have a service which resolves to a daemonset, and you want to check that the service won’t return an error if a node is taken down.
Scenario
scenarios:
- name: Check service responds when one node lost
steps:
# take down a single master VM
- nodeAction:
# pick all master VMs
matches:
- property:
name: "name"
value: ".*master.*"
filters:
# only pick the VMs which are in UP state
- property:
name: "state"
value: "UP"
# and only pick a single VM
- randomSample:
size: 1
actions:
- stop:
force: false
# make sure the service responds
- probeHTTP:
target:
service:
name: myservice
namespace: example
port: 8000
# assuming we had a 100 nodes, and it goes in round-robin
# we can make 100 calls to the service
count: 100