Migrating deployment processes and Continuous Integration at SAP SE
0 likes819 views
The document summarizes SAP SE's migration of their deployment processes and continuous integration to a more modern, future-proof system using tools like SLES12, Chef, GitHub, OBS, and KIWI. It overviews the software and processes used, including operating system image building with KIWI, configuration management with Chef, and version control with GitHub. The new system provides benefits like cleaner deployments, reproducibility, and maintainability compared to the previous process.
Ad
More Related Content
What's hot (20)
Viewers also liked (20)
Ad
Similar to Migrating deployment processes and Continuous Integration at SAP SE (20)
Ad
More from B1 Systems GmbH (20)
Recently uploaded (19)
Migrating deployment processes and Continuous Integration at SAP SE
- 1. Migrating deployment processes and Continuous Integration at SAP SE to a future-proof design using SLES12, Chef, GitHub, OBS and KIWI Florian Winkler Linux Consultant & Trainer B1 Systems GmbH [email protected] Eike Waldt Linux Consultant & Trainer B1 Systems GmbH [email protected]
- 3. 3 Introducing B1 Systems ● founded in 2004 ● operating both nationally & internationally ● ~ 100 employees ● vendor-independent (software & hardware) ● SUSE consulting & training partner ● focus: – consulting – support – training – development – operations – solutions
- 4. 4 Areas of expertise Monitoring NAGIOS RHEL Linux Operating Systems Systems Management Cloud Computing KVM XEN Virtualization ICINGA Configuration Management PUPPET SALT ANSIBLE RH SATELLITE SPACEWALK SUSE MANAGER SLES DEBIAN UBUNTU MIRANTIS SUSE CLOUD RH OSP OPENSTACK Virtualization
- 5. 5 Who we are ● Eike Waldt – Linux Consultant & Trainer – B1 employee since 05/2015 ● Florian Winkler – Linux Consultant & Trainer – B1 employee since 04/2014
- 6. 6 Close collaboration of ... – SaaS Cloud – HANA Enterprise Cloud – SAP Managed Service – SaaS Cloud Managed Services
- 7. 7 Agenda • Motivation and goals • Software used • Landscapes and processes • GMP - Global Management Portal • Operating system images • Deployment • Chef – Configuration management • Version control • Workflow • GMP - Integration of new tools • Chef design @ SAP’s clouds • Conclusion • Outlook
- 9. 9 Motivation and goals ● SLES12 SP1 at the doorstep – all the changes that come with it … ● historically grown deployment process – reinvent this completely – state of the art technology ● no real configuration management so far ● we wanted a fancy DevOps approach xD ● regaining a common base between several cloud landscapes
- 10. 10 Software used
- 11. 11 Software used ● SUSE Linux Enterprise Server 12 ● Open Build Service ● kiwi(-ng) ● jenkins ● GitHub Enterprise ● chef ● chefdk ● Berkshelf ● test-kitchen ● docker
- 13. 13 Admin Customer GMP Resource Pools SaaS HEC landscapes landscapes Bare Metal Hypervisor Xen KVM VMware Virtual Machines Hardware Pools VM Pools Landscape overview HEC 13 landscapes 13 Dcs 13.000 VMs 800 hypervisors 1.700 BareMetal SaaS 19 landscapes 8 DCs 32.000 VMs 3.700 hypervisors 1.100 Baremetal HEC SaaS
- 14. 14 GMP - Global Management Portal
- 15. 15 Global Management Portal (GMP) – network • network ranges • DHCP • DNS – hypervisor administration and deployment – virtual machine administration and deployment ● customer frontend ● inventory database – operating system images – monitoring – LDAP administration – storage administration – remote execution tool – auditing • SAP internal development • orchestration platform for ...
- 17. 17 OS images - ye olde way ● dump/restore a Goldmaster VM – update VM via “zypper up” – make changes directly in VM – different VMs for different image types • hypervisor • virtual machines Pitfalls • waste of resources • dumping a running system is unclean • no good version control/changelog • no reproducible builds • you do not really know what you are deploying
- 18. 18 OS images - the fancy way ● build images with kiwi in OBS – define content • in descriptive language • in file form – kiwi files and xml in git – minimalistic image approach Benefits • integration in already existing OBS • clean build – contents of the images are well- defined – reproducible builds • release management • ease of customization • great changelog in git • faster deployment • up-to-date packages
- 19. 19 Deployment
- 20. 20 Deployment - ye olde way 1) (PXE boot) 2) Restore dumped image 3) (Re)boot 4) Fetch/execute firstboot script a) perl scripts b) config file templates c) daemons to enable/start/disable/stop d) platform, DC specific parameters 5) Reboot 6) Machine is ready Pitfalls • data stored in files • unmaintainable • no “good” version control/changelog • no reproducible builds
- 21. 21 Deployment - the fancy way 1) (PXE boot) 2) Restore kiwi image 3) (Re)boot 4) cloud-init a) network settings b) chef-client settings 5) Execute chef-client a) config file templates b) daemons to enable/start/disable/stop c) platform, DC specific parameters 6) Reboot 7) Machine is ready Benefits ● clean deployment ● everything is kept in git ● great changelog in git ● way more maintainable
- 23. 23 Chef design basics ● client/server architecture ● stored data – attributes defined in roles • execution of code – cookbooks/recipes • design of dependencies – roles for landscapes/DCs (attributes) – nested roles for pools/customers/applications – run_list defines all roles and recipes for a client group or a single client • lifecycle and release management – environments define cookbook versions
- 25. 25 Version control - ye olde way and what doesn’t come with it ● local SVN repo for Perl scripts – only firstboot under control – no version control for external sources Pitfalls • no real AAA • no QA workflow • no approval workflow
- 26. 26 Version control - the fancy way and what comes with it ● git(hub) for almost everything – kiwi files – cookbooks • in their own separate repositories – chef environments Benefits • github is a well-known tool • using SAP’s internal github- enterprise • real AAA • Development /testing/ approvement/rollout workflows
- 27. 27 Workflow
- 28. 28 Workflow dev/test/deploy github jenkins chef-server GMP admin test environment docker/Xen hypervisor development workstation direct indirect chef-server
- 29. 29 GMP (Global Management Portal) – Integration of new tools
- 30. 30 Global Management Portal (GMP) – network • network ranges • DHCP • DNS – hypervisor administration and deployment – virtual machine administration and deployment ● customer frontend ● inventory database – operating system images – monitoring – LDAP administration – storage administration – remote execution tool – auditing • SAP internal development • orchestration platform for ...
- 31. 31 GMP integration ● automated chef-server install ● chef-server tasks – manipulating node objects • configuration database and mapping of entities – images to pools – chef-servers to pools – attributes to landscapes/pools – run_list to pools/nodes ● automated sync: GMP → (git) → chef-server – landscape attributes (roles) – environments, cookbooks and roles
- 32. 32 Chef design @ SAP’s clouds
- 33. 33 Chef design @ SAP’s clouds Admin Hardware Pools SaaS HEC run_list attributes role chef-server GMP Resource Pool: DC1-Customer 1 Resource Pool: DC2-Customer 2 Customer 1 environment VM Pools Hypervisor run_list Hypervisor run_list attributes role Bare Metal role run_list single attribute chef-server Customer 2 Landscape "SaaS 01" Landscape "SaaS 02" single attribute attributes role attributes base_role attributes bas_role run_list environment attributesattributes rolerole
- 34. 34 Conclusion
- 35. 35 Conclusion ● cleaned up deployment process ● working configuration management ● way better testing and integration workflow – implementing CI and CD ● identical infrastructure code in both SAP clouds ● future-proof – ready for new OS versions
- 36. 36 Outlook
- 37. 37 Outlook ● application deployment via Chef ● in-depth customization from customer side – cookbooks – coles – own chef servers (unmanaged) ● “backporting” deployment to SLES 11 ● establishing configuration management for SLES11
- 38. 38 Thank you
- 39. 39 Thank you for making this possible! ● Tools team – Sebastian Koehn – Sven Schubert ● HEC XEN team – Florian Kellmer – Ralf Lang – Christian Wolter ● MCD team – Tino Kaufmann – Alexander Lode ● SaaS Cloud team – Daniel Schier – Tobias Stolz – Ronny Tiebel • ticketweb – Sebastian Krieger
- 40. 40 Questions?
- 41. 41 Related Talks ... ● SUSE Manager 3 & SaltStack at Tyson Foods – Friday, Nov 11, 10:15 AM - 11:15 AM • The SUSE Manager Roadmap: A journey towards agile management of workloads in the enterprise – Friday, Nov 11, 9:00 AM - 10:00 AM