• OPENSHIFT , LINUX

OpenShift HA

#ha , #highly available

OpenShift HA

Good references and resources

DR options

Failure Scenarios

  • Complete datacenter outage

  • Availability Zone failure (network segment or group of racks)

  • Rack failure

  • Host failure

  • VM failure

  • OpenShift master failure

  • OpenShift etcd failure (etcd backup and restore)

Active - Active Production

  • deploy two active OpenShift clusters in separate datacenters

  • synchronize application rollouts between datacenters using CICD pipelines

  • spread application load balancing across two datacenters using a F5 GTM / Netscaler GSLB (Global Server Load Balancing, NetScaler Enterprise Edition.)

  • requires support from application’s backend datastore: i.e. replication/sharding/etc across datacenters

  • optionally, deploy backend datastore in one datacenter and plan for its failover separately from the application (e.g. front end app uses backend Oracle database, which handles its own sync/replication/restore): this results in active/active on the frontend with active/passive on the backend, where one active datacenter will always be configured to handle database requests to the single datastore in the other datacenter

  • you could choose to handle applications individually, and any applications that do not require persistent data can always run active/active while those that do require persistent data run active/passive with managed replication

Active - Passive Production

  • deploy a secondary passive cluster is a separate datacenter from the active OpenShift cluster

  • synchronize application rollouts between datacenters using CICD pipelines

  • employ a load balancer to switch to secondary passive cluster

  • replicate application data from active to passive cluster on a reasonable interval

Active Production - Active Nonprod Staging Standby

  • deploy Production and Nonproduction in separate datacenters

  • use Nonproduction isolated nodes as a Production DR

  • switch load balancers for applications from Prod to Nonprod DR

  • already have environment ready, just need to synchronize all deployments to latest prod releases