I would like to try set up Kubernetes cluster in AWS for an application that consists of:
AWS has been chosen instead of GCE, because other services, e.g. S3, are already being used. Currently the app is set up using Jelastic (https://jelastic.com/ - GUI for cluster management), however Jelastic ceased to suffice because of lack of automation tooling and also hardware provider had numerous outages.
Jelastic uses cloudlets as an abstraction of computer resources - one cloudlet is 200MHz CPU and 128MB RAM. The app uses about 150 cloudlets = 30 GHz CPU + 20GB RAM.
How would you recommend setting up the cluster in AWS? What instance types are the best for Kubernetes master and minions? Do you recommend running databases on the cluster as well as services, or is it better to spin up dedicated instances for them?
To fine tune aws instance type requires a bit of trial & error ... for starters if you define these env vars prior to launching kube-up.sh
export MASTER_SIZE=t2.large # master instance type
export NODE_SIZE=t2.large # minion instance type
export NUM_NODES=3 # determines number of minion instancesthen once running closely read over output to call
kubectl describe nodes ... typical output
Namespace Name CPU Requests CPU Limits Memory Requests Memory Limits
--------- ---- ------------ ---------- --------------- -------------
kube-system elasticsearch-logging-v1-rvvnw 100m (5%) 100m (5%) 0 (0%) 0 (0%)
kube-system elasticsearch-logging-v1-ybz5n 100m (5%) 100m (5%) 0 (0%) 0 (0%)
kube-system fluentd-elasticsearch-ip-172-20-0-211.ec2.internal 100m (5%) 0 (0%) 200Mi (5%) 200Mi (5%)
kube-system heapster-v1.1.0-1261016182-q18fr 288m (14%) 288m (14%) 598916Ki (14%) 598916Ki (14%)
kube-system kibana-logging-v1-9337e 100m (5%) 100m (5%) 0 (0%) 0 (0%)
kube-system kube-dns-v17-iits3 110m (5%) 110m (5%) 120Mi (3%) 220Mi (5%)
kube-system kube-proxy-ip-172-20-0-211.ec2.internal 100m (5%) 0 (0%) 0 (0%) 0 (0%)
kube-system kubernetes-dashboard-v1.1.0-ss59o 100m (5%) 100m (5%) 50Mi (1%) 50Mi (1%)
kube-system monitoring-influxdb-grafana-v3-lzjq7 200m (10%) 200m (10%) 600Mi (15%) 600Mi (15%)
ruptureofthemundaneplane loudspeed-deployment-3003771997-bybrm 0 (0%) 0 (0%) 0 (0%) 0 (0%)
ruptureofthemundaneplane mongo-deployment-2286179029-0bltg 0 (0%) 0 (0%) 0 (0%) 0 (0%)
Allocated resources:
(Total limits may be over 100%, i.e., overcommitted. More info: http://releases.k8s.io/HEAD/docs/user-guide/compute-resources.md)
CPU Requests CPU Limits Memory Requests Memory Limits
------------ ---------- --------------- -------------
1198m (59%) 998m (49%) 1592196Ki (39%) 1694596Ki (41%)in above focus on that bottom set of resource usage % values
https://aws.amazon.com/ec2/instance-types/
https://aws.amazon.com/ec2/pricing/
tweak to achieve smallest/cheapest instance types without getting too high usage % values
Can you please give us more information about your use case and automation you want to have? In Jelastic there is an automation tool called CloudScripting, with a help of this any kind of automation is possible.