Continuous Delivery of Stateful Applications with Kubernetes (in Production)
2 likes291 views
The document discusses the continuous delivery of stateful applications using Kubernetes, emphasizing the importance of container management and automation in production environments. Key topics include best practices for stateful apps, data management challenges, and a solution provided by Portworx for persistent storage. The presentation highlights the integration of Codefresh and Portworx in simplifying CI/CD processes while ensuring data integrity and availability.
Continuous Delivery of Stateful Applications with Kubernetes (in Production)
- 1. Continuous Delivery of Applications with Kubernetes in Prod Dan Garfield Chief Evangelist Codefresh Eric Han Product Manager Portworx Sta l ^ April 25, 2018
- 2. Dan Garfield Chief Evangelist, Codefresh Customers Over 3.5 million images built
- 3. 3© 2018 PORTWORX with CODEFRESH Eric Han Product Management, Portworx Customers Trusted in production by Fortune Global 500 and other leading companies.
- 4. What we’ll cover today ● Intro to Stateful Apps ● Best Practices ● Functional Testing w/ Data ● Continuous Delivery w/ Data ● Demos + Q&A!
- 5. 5© 2018 PORTWORX with CODEFRESH Every application requires Stateful containers Linux distros NoSQL RDBMS Content & Builds Analytics & Processing Web Stateful Stateless
- 6. 6© 2018 PORTWORX with CODEFRESH Before Containers Single application per server Manual CI/CD processes are common array x86x86x86 Business Tier Ridgid deployments, scales poorly High costs per application
- 7. 7© 2018 PORTWORX with CODEFRESH With Containers Dynamic app demand / mix Dense (20+) apps per server x86x86x86 Deployment, mgmt becomes automated Lower costs per application Sta l ^
- 8. 8© 2018 PORTWORX with CODEFRESH Persistence + Portability Storage Class Highly Available Persistent Volume postgres production Persistent Volume Claim postgres prod volume Pod mongo Storage Class encrypted + HA Persistent Volume postgres staging Portable applications + leveraging infrastructure
- 9. 9© 2018 PORTWORX with CODEFRESH Pod to Persistent Volume kind: StorageClass metadata: name: px-ha provisioner: portworx parameters: fs: “ext4” io_priority: “high” repl: “3” Storage Class kind: PersistentVolume metadata: name: px-disk-1 spec: capacity: storage: 8Gi Persistent Volume kind: PersistentVolumeClaim metadata: name: seeking-volume spec: resources: requests: storage: 8Gi storageClassName: px-ha selector: matchLabels: release: stable Persistent Volume Claim kind: Pod metadata: name: mypod spec: containers: - name: database image:postgres volumeMounts: - mountPath: /var/lib/pg/data volumes: - name: pg-data persistentVolumeClaim: claimName: seeking-volume Pod 1 4 2 Portable applications + leveraging infrastructure
- 10. 10© 2018 PORTWORX with CODEFRESH Top Problems with Stateful Containers in Production 1. Stuck without HA: app wants to failover to new server but data is stuck/mounted to EBS/drive on the old server 2. Data protection: snapshot, back-up, and encryption are at the VM-level and not at the app level 3. No Performance Control: running multiple containers per server = poor performance EBSEBS EBS x86x86x86
- 11. 11© 2018 PORTWORX with CODEFRESH More complexity is not the answer ▪ Stateful containers? ▪ Persistent Volumes Claims? ▪ Servers with dozens of apps?
- 12. 12© 2018 PORTWORX with CODEFRESH Portworx is a container Software-Defined-Storage … x86x86x86 x86 ▪ Elasticity: aggregate storage capacity and enables storage-less servers to participate ▪ Container-granular HA: each storage volume is per container; Portworx replicates to a quorum ▪ Control: unifies storage control of 1000 servers on-prem and in-cloud SSD HDD SSD HDD SSD HDD SSD HDD … 1000 servers
- 13. 13© 2018 PORTWORX with CODEFRESH that orchestrates scheduling with storage … x86x86x86 x86 ▪ Placement: ensures optimal compute scheduling with storage ▪ Seamless: integrates the user experience with schedulers (data placement, rack info, media health)
- 14. 14© 2018 PORTWORX with CODEFRESH and manages the full data lifecycle. x86x86x86 x86 ▪ BYOK Encryption: integrates with Vault, AWS KMS, and others to protect data in-flight and at rest ▪ Backup/Restore: to another cluster or off-prem object storage S3 object store Snap Restore
- 15. 15© 2018 PORTWORX with CODEFRESH Codefresh and Portworx Continuous Delivery ID1 triggered snapshot ID2 Portworx Codefresh pipeline Production cluster prod staging
- 19. Continuous Delivery Pipeline w/Live Data Commit Build images/Chart Deploy to one-off environment Clone live data Functional Tests Deploy to Production Remove one-off env
- 20. Demotime!
- 21. Try Portworx on GKE Contact us at: portworx.com Schedule a 1:1 with us Get in touch at: Codefresh.io Q T !