OCI Foundations - Compute: Difference between revisions

From Smithnet Wiki
Jump to navigation Jump to search
m (10 revisions imported)
 
(2 intermediate revisions by the same user not shown)
Line 1: Line 1:


== Compute Options ==
== Introduction ==


* Virtual Machines / Bare Metal servers / Dedicated Host (will run ONLY your VMs, not shared with other customers).
* Virtual Machines / Bare Metal servers / Dedicated Host (will run ONLY your VMs, not shared with other customers).
Line 41: Line 41:


Fully managed, scalable, highly available based on Kubernetes
Fully managed, scalable, highly available based on Kubernetes
Worker Node: machine on which Kubernetes is installed
* Worker Node: machine on which Kubernetes is installed
Node Pool: group of Nodes
* Node Pool: group of Nodes
Pod: Group of one of more containers in a cluster
* Pod: Group of one of more containers in a cluster


Control Plane nodes:
Control Plane nodes:

Latest revision as of 04:53, 8 May 2024

Introduction

  • Virtual Machines / Bare Metal servers / Dedicated Host (will run ONLY your VMs, not shared with other customers).
  • Can choose Intel, AMD or ARM.
  • Preemptible VMs - short term batch jobs, lower cost.

Flexible Shape: Choose cores and RAM.

Instance Basics

Dependencies:

  • A VCN is required
  • The host's VNIC is placed in the VCN
  • Boot image and data on remote storage
  • Connected to Block Storage

Live migrate: transparent move on a failure to another host, applications keep running

Use Cloud Shell to generate SSH keys: ssh-keygen -b 2048 -t rsa -d test_key

Create instance: upload public key to allow SSH login: ssh -l test_key opc@<ip>

Scaling

Vertical Scaling: up/down shapes (CPUs, RAM)

  • Downtime is required
  • Best to stop instance first

Horizontal Scaling:

  • Add/remove more VMs of the same shape
  • Adds reduncancy and capability for load peaks

Autoscaling

  • Running instance -> Config -> Instance Pool
  • Scaling Rule for min/max size, thresholds for scaling

Oracle Container for Kubernetes (OKE) Cluster

Fully managed, scalable, highly available based on Kubernetes

  • Worker Node: machine on which Kubernetes is installed
  • Node Pool: group of Nodes
  • Pod: Group of one of more containers in a cluster

Control Plane nodes:

  • manage the worker node and the Pods in the cluster
  • etcd: key/value pair database
  • Managed by Oracle

Enhanced CLusters:

  • Support all features
  • Finanically backed SLAs

Basic Clusters:

  • Basic features
  • Non-finanical SLAs

Virtual Node:

  • Serverless option
  • Kubernetes is upgraded and security patches are applied while respecting application availability
  • Managed by Oracle
  • Can only be created in Enhanced Clusters

Managed Nodes

  • Customer is responsible for managing the nodes
  • Customer is responsible for upgrading Kubernetes and the capacity
  • Can be created in Basic and Enhanced Clusters

Container Workloads

Container Instances: A simple way to run a container application without the complexity of Kubernetes

  • Users supply the container image and resource limits for containers
  • Serverless offering

Oracle Functions

Write code in a runtime and the cloud provider executes the code. Not even a container is needed.

  • Function as a Service
  • Event Driven Architecture
  • Billed only for when the function runs
  • Powered by OpenFn

Process:

  • Push container to registry
  • Configure trigger function (CLI, API, OCI events)
  • Code only runs when triggered