GKE Cheatsheet

Basics 

------ 

 

Installing gcloud

curl https://sdk.cloud.google.com | bash  exec -l $SHELL

 

Compute Instances 

----------------- 

 

Auto scaling	- create image  - create instance template  - create instance group (tune auto scaling)
Spinning Up compute instance via ansible	1. Create credentials file  * see "Generatin credentials file" under IAM/Security    2. Install pip packages  pip install apache-libcloud PyCrypto    3. Create a simple playbook  ---  - hosts: localhost    connection: local    gather_facts: False    tasks:      - name: Create VM instance      gce:        instance_names: my-test-instance1        zone: us-central1-a        machine_type: n1-standard-1        image: debian-8        state: present        service_account_email: "639770814480-compute@developer.gserviceaccount.com"        credentials_file: "~/Downloads/creds-my-private-cloud-01.json"        project_id: "my-private-cloud-01"        disk_size: 10    4. Run playbook  ansible-playbook create_gcp_instance.yml
Moving an instance to a different zone within same region	NOTE:    If there are errors regarding reaching static address limit, try to delete some  static IPs first.    Move can also be done while instance is online.    gcloud compute instances move instance-1 --destination-zone=asia-east1-a
SSH'ing using from local laptop to target using internal IP	gcloud compute ssh --zone asia-southeast1-a <compute instance name> --internal-ip

 

IAM/Security 

------------ 

 

Generating credentials file

- go to: APIs & Services > Credentials  - choose: Create credentials > Service account key    * Compute Engine default service account    * JSON format

 

GKE 

--- 

 

Authenticating to cluster	gcloud auth activate-service-account --key-file=creds.json  gcloud container clusters get-credentials my-sample-cluster --zone europe-west2-c
Adding insecure registry	1. Update DOCKER_OPTS under /etc/default/docker of each nodes to look like this:  DOCKER_OPTS="-p /var/run/docker.pid [...] --insecure-registry docker.registry2:5000 --insecure-registry docker.registry2:5001"    2. Drain the node.  kubectl drain node1    3. Restart docker.  systemctl restart docker    4. Repeat steps 1 -3 on the remaining nodes.    https://stackoverflow.com/questions/49206139/what-is-the-proper-way-of-adding-insecure-registry-in-google-kubernetes-engine-n
Scenarios in draining a node	Scenario 1:  You drain a node and remove that from a 3 node instance group.. Instance group scaled down to 2 node.  You update instance group to 3 nodes. So there are now 4 nodes (1 is drained as seen from kubectl).  You want to remove that drained node. You can simply delete the compute instance directly and wait  for few minutes. The next time you hit "kubectl get nodes", there will now be only 3 nodes that are  ready.