Schedule legacy applications as single instance on Kubernetes -


a lot of legacy applications deployed containers. of them need few changes work in container many of them not built scale, example because maintain session data or write volume (concurrency issues).

i wondering if applications intended run on kubernetes , if way so. pods not durable, desired way start application using replication controller , setting replicas 1. rc ensures right amount of pods running. documentation specifies kills pods if there many. wondering if that's ever case (if pod not started manually).

i guess database postgres (with external data volume) example. have seen tutorials deploying using replication controller.

creating replication controller 1 replica indeed approach, it's more reliable starting single pod since benefit auto-healing mechanism: in case node app running on dies, pod terminated restarted somewhere else.


data persistence in context of cluster management system kubernetes means data should available outside cluster itself (separate storage). use ec2 ebs since our app runs in aws, kubernetes supports lot of other volume types. if pod runs on node a, volumes uses mounted locally , inside pod containers. if pod destroyed , restarted on node b volume unmounted node a , mounted on node b before containers of pod recreated. pretty neat.

take @ persistent volumes, should particularly interesting you.


Comments