inwinSTACK - ceph integrate with kubernetes
2 likes954 views
This document summarizes a presentation about integrating Ceph with Kubernetes. It discusses key updates in the Ceph Luminous release, including Bluestore, erasure coding overwrite, multiple MDS, and the Ceph MGR dashboard. It then explains why Kubernetes and Ceph are good choices for storage. Finally, it covers different approaches to integrating Ceph and Kubernetes, such as using CephFS or CephRBD, and hyperconverged vs hyperscale architectures.
inwinSTACK - ceph integrate with kubernetes
- 1. Ceph integrate with Kubernetes Speaker : Thor Chin
- 2. Agenda Ceph Luminous important Features update Why Kubernetes and Ceph Ceph integrate with Kubernetes Ceph integrate with Kubernetes
- 3. Ceph Luminous important Features update
- 5. Ceph Luminous important Features update ※ Bluestore ※ Erasure Coding Overwrite ※ Multiple MDS ※ Ceph MGR Dashboard
- 6. Bluestore
- 7. FileStore vs BlueStore FileStore BlueStore Filesystem Device Journal Direct Write / WAL Disk Disk, SSD, NVME
- 9. Replica vs Erasure Coding Replica Erasure Coding space used 3x (n=3) 1.5x (k+m=3) CPU overhead Normal High support RBD, RGW, CephFS RGW only
- 10. Replica vs EC Overwrite Replica EC Overwrite space used 3x (n=3) 1.50x (k+m=3) CPU overhead Normal High support ALL ALL
- 11. Single MDS Ceph Storage Cluster MDS. 0 MDS. 1 MDS. 2 MDS. 3 Active Standby
- 12. Multiple MDS Ceph Storage Cluster MDS. 0 MDS. 1 MDS. 2 MDS. 3 Active Standby
- 13. Directory Fragmentation M M M M M M M M M M M M M M M M M M M M M M M M MDS.0 MDS.1MDS.0
- 14. Ceph MGR Dashboard Architecture MGRDashboard
- 15. Ceph Health
- 16. Pool Usage
- 17. Host Monitor
- 18. Logging
- 19. CephFS Monitor
- 20. Ceph Monitoring Tools Comparison Calamari Grafana Nagios OpenATTIC Ceph MGR Dashboard Supported Version Hammer ALL ALL ALL Luminous Cluster Status ✓ ✓ ✓ ✓ ✓ Pool Status ✓ ✓ ✓ ✓ ✓ Pool Mgmt ✓ ✓ OSD Status ✓ ✓ ✓ ✓ ✓ OSD Mgmt ✓ IOPS / Throughput ✓ ✓ ✓ MDS Status ✓ ✓ ✓ CephFS Status ✓ Extra Customized and Time Range Email Alert
- 21. Why Kubernetes and Ceph
- 22. Why Kubernetes? Powerful Community Open Source – no vendor lock-in Large Scale Rolling Upgrade Self-healing Federation and Hybrid Easy to Management Easy to integrate with monitoring system – Heapster, InfluxDB, Grafana
- 23. Why Ceph ? Distributed storage system Fault tolerant , no SPoF X86 Commodity hardware Saving you costs, giving you flexibility Large scale – Incremental expansion 10s to 1000s of nodes Unified storage platform Scalable object , Block , File system Open source – No vendor lock-in Automatically balance the file system Data security with 2 or more copies in different physical store media 23
- 25. CephFS or CephRBD? CephFS with K8S Ceph RBD with K8S Multiple Read Yes Yes Multiple Write Yes No Quota No Yes (Image Size) HA support Yes Yes Performance Good enough Better Isolated volume No (shared) Yes
- 26. Hyper Converged VS Hyper Scale Architecture
- 28. Hyper Converged and Containerized Architecture K8S Ceph-Mon Ceph-MDS Ceph-OSD master K8S Ceph-Mon Ceph-MDS Ceph-OSD K8S Ceph-Mon Ceph-MDS Ceph-OSD 10g master master Admin node Auto- discovery Auto- deploy
- 29. Mixed Architecture K8S Ceph-Mon Ceph-MDS Ceph-OSD master K8S Ceph-Mon Ceph-MDS Ceph-OSD K8S Ceph-Mon Ceph-MDS Ceph-OSD 10g master master Admin node Auto- discovery Auto- deploy Storage node ceph-osd ceph-osd ceph-osd Storage node ceph-osd ceph-osd ceph-osd Storage node ceph-osd ceph-osd ceph-osd K8S node POD A POD B POD C K8S node POD A POD B POD C
- 30. www.inwinstack.com Thank You! 迎 棧 科 技 股 份 有 限 公 司