|
|
@@ -1,37 +1,5 @@
|
|
|
---
|
|
|
# Prepares the environment for the apps-review exercise, and cleans up afterwards.
|
|
|
-#
|
|
|
-# TODO: create some projects:
|
|
|
-# - one with a project node selector
|
|
|
-# and a deployment with conflicting node {selector|affinity} ???
|
|
|
-# - one with a very low quota
|
|
|
-# and a deployment which exceeds the quota
|
|
|
-# - taint a node
|
|
|
-# and select a deployment to run a pod on it
|
|
|
-# then debug these conditions and fix them
|
|
|
-#
|
|
|
-# TODO: make two nodes unschedulable, create a project, and deploy an application, scale to three
|
|
|
-# make the nodes schedulable again, use podAntiAffinity to disperse the pods, scale to 6 and see scheduling
|
|
|
-#
|
|
|
-# simulate load (loadtest? loadgenerator?) beyond container's cpu limit and then improve performance by raising limit
|
|
|
-#
|
|
|
-# TODO: probes with extremely low cpu limit, see them crashloop, fix it
|
|
|
-#
|
|
|
-# use stress-ng ap to allocate all memory (more than limit), monitor the metrics to diagnose the crash
|
|
|
-#
|
|
|
-# client-server apps, low limits, monitor performance
|
|
|
-#
|
|
|
-# custom metrics, grafana
|
|
|
-#
|
|
|
-# TODO: run two instances on the same node, no pdb, drain the node - observe failure in another terminal
|
|
|
-# repeat with pdb, see no failures
|
|
|
-#
|
|
|
-# recreate strategy, rollout a change, observe outage in another terminal
|
|
|
-# switch to rolling w/maxUnavailable, repeat, see no failures
|
|
|
-#
|
|
|
-# deploy an app w/requests, generate load, observe timing
|
|
|
-# add HPA, generate load, compare
|
|
|
-#
|
|
|
- name: Prepare the exercise of apps-review.
|
|
|
hosts: localhost
|
|
|
gather_subset: min
|
