Deploying and operating GeoServer: a DevOps perspective (FOSS4G 2022 Edition)

Cloud computing is revolutionizing the way companies develop, deploy and operate software and GeoSpatial software is no exception. With benefits of cloud based deployments range from cost savings to simplified management, flexibility, lower downtime and scalability of dynamic environments it is easy to understand why more and more companies are migrating their on premise systems to the cloud but cloud based setups have their own set of hurdles and challenges.
The migration of the series itself can be challenging. Monitoring, debugging and scaling of applications are very much different than what you are used to.
In this presentation we will share with you the lessons we have learned at GeoSolutions to tackle these problems and share some common patterns for the migration of on premise GeoServer clusters to the cloud. We'll share with you tips on how to:

best practices to migrate your existing GeoServer cluster to the cloud
insights on your geoserver cluster using centralized logging and Monitor plugin
avoid common bottlenecks to best set up a distributed scalable GeoServer cluster 
work containers and container orchestrators like Kubernetes

Transcript

1. Alessandro Parma
   Luca Pasquali
   GeoSolutions
   Deploying and operating GeoServer:
   a DevOps perspective
   

   View Slide

2. Introduction
   

   View Slide

3. GeoSolutions
   Enterprise Support
   Services
   Deployment
   Subscription
   Professional
   Training
   Customized
   Solutions
   GeoNode
   • Offices in Italy & US, Global Clients/Team
   • 40+ collaborators, 30+ Engineers
   • Our products
   • Our Offer
   

   View Slide

4. Affiliations
   We strongly support Open
   Source, it Is in our core
   We actively participate in
   OGC working groups and
   get funded to advance new
   open standards
   We support standards
   critical to GEOINT
   

   View Slide

5. What is GeoServer?
   (really??)
   

   View Slide

6. What is GeoServer?
   • GeoSpatial enterprise gateway
   • Manage and Disseminate of raster and vector data
   • Standards compliant
   • OGC WCS 1.0, 1.1.1 (RI), 2.0
   • OGC WFS 1.0, 1.1 (RI), 2.0
   • OGC WMS 1.1.1, 1.3.0
   • OGC WPS 1.0.0
   • OGC CSW 2.0.1 (ebRIM)
   • Google Earth/Maps support
   • KML, GeoSearch, etc..
   

   View Slide

7. What is GeoServer?
   GeoServer
   WFS
   WMS
   PostGIS
   Oracle
   H2
   DB2
   SQL
   Server
   GeoPacka
   ge
   MySql
   Spatialite
   Elastic
   MongoDB
   Shapefile ----------
   ----------
   ---------
   ----------
   ----------
   ----------
   ---------
   ----------
   ----------
   ----------
   ---------
   ----------
   WFS
   PNG, GIF
   JPEG
   TIFF,
   GeoTIFF
   SVG, PDF
   KML/KMZ
   Shapefile
   GML2
   GML3
   GeoRSS
   GeoJSON
   CSV/XLS
   GeoPackage
   Raw vector
   data
   Servers
   Styled
   maps
   DBMS
   Vector files
   WCS
   GeoTIFF
   WMS
   ArcGrid
   Img+world
   Mosaic
   MrSID
   JPEG 2000
   ECW,Pyramid,
   Oracle
   GeoRaster,
   PostGis
   Raster files
   Raw raster
   data
   GeoTIFF
   ArcGrid
   GTopo30
   Img+World
   WMTS,
   TMS
   KML superoverlays
   Google maps tiles
   OGC tiles
   OSGEO tiles
   KML
   WPS
   CSW
   ESRI
   REST
   

   View Slide

8. Applications deployment
   

   View Slide

9. App. Execution Technologies
   • How things changed over time
   • Bare Metal (non virtualized)
   Buy or Rent the HW and install everything from scratch. Single
   Tenant
   • Virtual Machines
   Spin up a virtual server along with others running on an existing,
   virtualized server. Looks like a dedicated machine
   • Containers
   Run a single (?) application alongside others on an existing
   environment
   T
   I
   M
   E
   

   View Slide

10. Bare Metal (no virt.)
    • Pros
    • Fast (no virt. layer)
    • Single tenant -> Secure
    • No “noisy neighbour” effect -> more stability
    • Cons
    • Limited flexibility & scalability
    • billing, re-installs
    • Dimensioning is hard -> over-provision
    • More work for Operations
    

    View Slide

11. Virtual Machines
    • Pros
    • Flexibility (Start, Stop, Snapshot, Template)
    • Fragmentation and resource usage
    • Reduces costs
    • Cons
    • Virtualization layer overhead
    • Noisy neighbour effect
    • Single failure impacts many services
    

    View Slide

12. What is a Container
    •
    What is a container then?
    •
    Type of virtualization that happens at the
    operating system level
    •
    Applications can run in an isolated user
    spaces called “Containers”
    •
    Implemented at the kernel level, multiple
    containers share the same Kernel
    

    View Slide

13. Containers vs VMs
    ●
    How does it compare to VMs?
    

    View Slide

14. Containers
    • Pros
    • Application is bundled with dependencies
    • Shared Kernel (lower resource)
    • Easy installation and migration
    • High density with isolation
    • Startup time
    • Cons
    • Steep learning curve
    • Existing apps must be “containerized” first
    • Shared Kernel (security?)
    

    View Slide

15. Kubernetes
    

    View Slide

16. ●
    Platform to manage containers and services
    ●
    Originally Developed by Google based on
    their experience with Borg
    ●
    Manage a set of nodes to provide you with a
    platform to run the containers
    ●
    Helps you manage and scale your applications
    What is Kubernetes
    

    View Slide

17. Why is it relevant?
    ●
    Traditional deployments
    ●
    No resource boundaries → some applications
    starve for resources
    ●
    Can’t easily reallocate resources after the initial
    setup
    ●
    Virtual Machines
    ●
    Multiple VMs on the same server → better
    resource utilization
    ●
    Better isolation
    ●
    Each VM has a copy of the OS
    

    View Slide

18. Why is it relevant?
    ●
    Containers
    ●
    shared kernel with isolated userspace
    ●
    each container has its own filesystem, a share of
    CPU cores
    ●
    decoupled from they underlying infrastructure →
    portable across distributions and cloud providers
    ●
    …
    

    View Slide

19. Why is it relevant?
    ●
    …
    ●
    fast image creation and easy rollback
    compared to VMs → Good fit for frequent
    deployments and CI/CD
    ●
    separation of concerns between Devs and
    Ops
    ●
    consistency across development in multiple
    environments
    

    View Slide

20. Why Kubernetes?
    

    View Slide

21. Why Kubernetes?
    ●
    Manage the containers that run your
    applications in production with no downtime
    ●
    Takes care of running your application
    containers on a distributed system
    ●
    Scaling application and the nodes cluster and
    failover
    

    View Slide

22. ●
    Also
    ●
    Service discovery and load balancing
    ●
    Storage Abstraction mount storage of
    choice)
    ●
    Auto rollouts and rollbacks. You describe
    the desired state for your containers
    ●
    Self Healing with restart failing containers,
    and probes, ..
    ●
    Secrets management. Externalize sensitive
    and env. specific info
    Why Kubernetes?
    

    View Slide

23. How does it compare to..
    ●
    There are other orchestrators and tools
    available to manage containers
    ●
    Docker Compose
    ●
    allows you to define services as collections but
    that is pretty much it
    ●
    Docker Swarm
    ●
    gives you to work on a distributed environment
    ●
    services definition and commands are somewhat
    similar to compose
    ●
    not as sophisticated (and complex! as K8s)
    

    View Slide

24. Containerize GeoServer
    ●
    Need a docker image
    ●
    Use the ones available!
    ●
    GeoSolutions here, sources here
    ●
    Official Image available check blogpost here
    ●
    Ready to use
    ●
    Based on Tomcat images
    ●
    Configurable
    

    View Slide

25. Example K8s deployment
    

    View Slide

26. Helm
    • Package Manager for Kubernetes
    • Use “recipes” available online
    • Deployment unit is called a Chart
    • Charts can easily be published, versioned and
    shared
    • Provide templating for your Manifests
    • GeoSolutions Chart implementation available for
    preview:
    https://github.com/geosolutions-it/charts
    

    View Slide

27. Resources
    ●
    What is Kubernetes
    ●
    Borg: The Predecessor to Kubernetes
    ●
    Containerization
    ●
    What is a Container
    ●
    Docker Compose
    ●
    Docker Swarm
    ●
    Rancher
    

    View Slide

28. Best Practices
    

    View Slide

29. Best Practices …
    • Start simple
    • Set up a local environment (Minikube)
    • Use available Images and Charts, don’t reinvent the wheel
    • Use managed K8s
    • Plan ahead
    • Design your infrastructure before moving on with
    deployment
    • Clustering Strategies, Storage Classes , Volumes
    sizing
    • Implement a Logging and a monitoring solution
    

    View Slide

30. … Best Practices
    • Choose the right nodes
    • No need for tons of RAM
    • Compute optimized node perform better (~4
    CPUs/instance)
    

    View Slide

31. Storage Classes
    

    View Slide

32. Choose Storage - Block Storage
    • Local or Block Storage
    • Local
    • Non Shared
    • Can fail
    • Low latency
    • Good fit for temporary
    storing Logs and Audits
    • Cached Tiles?
    

    View Slide

33. Choose Storage - File Share
    • File Share
    • File Storage ~ NFS
    • Shared (locally)
    • Scales well
    • Typical use case is GeoServer
    datadirs
    • Spatial Data
    • Cached tiles?
    

    View Slide

34. Choose Storage - Blob Storage
    • Blob Storage
    • Max scalability
    • Elastic
    • Cheap
    • High Latency
    • Shared
    • Robust
    • Cached Tiles?
    

    View Slide

35. Use Cases
    

    View Slide

36. EO Data Dissemination
    ●
    Earth Observation
    ●
    Meteorological and Oceanographic data
    ●
    Continuous data ingestion flordata
    ●
    TBs of data growing over time. Raster and Vector
    ●
    

    View Slide

37. EO Data Dissemination
    

    View Slide

38. Observability and Debugging
    

    View Slide

39. Monitoring and Logging
    • Can be tricky in Containerized
    environments!
    • Dynamic and distributed
    • Instances are spinning
    up and down
    • distributed across multiple
    nodes
    • Hard to identify problems
    • What can we do?
    

    View Slide

40. Centralize and Aggregate
    • Centralize and Aggregate
    • Single central location
    • Easy to navigate and filter
    • New nodes can spawn but also go away
    • You’ll need “shippers” to collect and
    send out logs to the central service
    

    View Slide

41. Collect Metrics
    • Collect Metrics
    • Resp Time,
    • Throughput,
    • Uptime,
    • Error Rate
    • ...
    

    View Slide

42. Auditing
    • GeoServer can produce audit files for
    you
    • Monitoring extension
    • Tracks requests made to GeoServer
    • Collect and ingest them to create pretty
    Dashboards
    

    View Slide

43. Auditing
    • Audit Event
    

    View Slide

44. Analyze performance of your Cluster
    

    View Slide

45. Analyze performance of your Cluster
    

    View Slide

46. Pointers
    • GeoSolutions Website
    https://geosolutionsgroup.com/
    • GeoServer on Kubernetes
    https://www.geosolutionsgroup.com/blog/devops-k8
    s/
    • A DevOps perspective on GeoServer
    https://www.geosolutionsgroup.com/blog/devops-geoser
    ver-monitoring-metering/
    • Cloud Optimized GeoTiffs
    https://www.cogeo.org/
    https://docs.geoserver.org/latest/en/user/community
    /cog/index.html
    • Helm
    https://helm.sh/
    

    View Slide

47. The End
    Questions?
    alessandro.parma@geosolutionsgroup.com
    luca.pasquali@geosolutionsgroup.com
    info@geosolutionsgroup.com
    

    View Slide