Application Container Design

Selbst wenn man gemäß Domain-Driven Design passende vertikale Schnitte der Applikation gefunden hat und sie mittels einer Microservice-Architektur umsetzen will, sind noch immer genügend Fragen offen: Aus wie vielen Containern besteht mein Service? Wie modularisiere ich sinnvoll? Wie mache ich die Konfiguration? Wie die Kommunikation? Wie skaliere ich am besten? In seinem Talk stellt Christopher Schmidt die Best Practices der Kubernetes- und Container-Pattern vor und zeigt am Beispiel Strimzi (Kafka Operator) oder dem Service Mesh Istio, wie sie sich erfolgreich umsetzen lassen.

Transcript

1. INNOQ / Application Container Design mit Kubernetes INNOQ Technology Night Zürich Application Container Design mit Kubernetes
2. https://innoq.com/ Berlin ・ Düsseldorf ・ Frankfurt ・ München ・ Zürich 2
3. Berlin ・ Düsseldorf ・ Frankfurt ・ München ・ Zürich Christopher Schmidt Senior Consultant @fakod • Kubernetes • Container • Architecture • Training • Consulting 3
4. INNOQ / Application Container Design mit Kubernetes From DDD to Production Architecture ❏ SCS ❏ µServices DDD ❏ Contexts ❏ Ctx Mapping 4
5. INNOQ / Application Container Design mit Kubernetes Recap Microservices ❏ easy to replace ❏ one team, one µService ❏ small in size ❏ messaging enabled ❏ bounded by contexts ❏ independently deployable ❏ implemented using different technologies 5
6. INNOQ / Application Container Design mit Kubernetes Recap SCS Systems as complete units of DB, Logic, UI Isolated, independent, autonomous System System 6
7. INNOQ / Application Container Design mit Kubernetes From DDD to Production ? Production ❏ Container ❏ Platform Architecture ❏ SCS ❏ µServices DDD ❏ Contexts ❏ Ctx Mapping 7
8. INNOQ / Application Container Design mit Kubernetes Kubernetes Capabilities Node 1 Node 2 Node 3 Node 10000 Master Master Master 110 110 110 110 ... 8
9. INNOQ / Application Container Design mit Kubernetes My Application Spring Boot Provide production-ready features such as ❏ metrics, ❏ health checks ❏ configuration Stand-alone application Embed Tomcat, Jetty Automatically configures Spring Jobs Spring MVC 9
10. INNOQ / Application Container Design mit Kubernetes Finally: My Application Spring Boot can do everything anyway DB solves any persistence issues 10
11. INNOQ / Application Container Design mit Kubernetes Kubernetes and Application Node 1 Node 2 Node 3 Node 10000 Master Master Master ... 11
12. INNOQ / Application Container Design mit Kubernetes Application Container Design Production ❏ Container ❏ Platform Architecture ❏ SCS ❏ µServices DDD ❏ Contexts ❏ Ctx Mapping Application Container Design ? 12
13. INNOQ / Application Container Design mit Kubernetes there is no software architecture without production view 13
14. INNOQ / Application Container Design mit Kubernetes 4+1 architectural view model Logical view: • functionality provided to end-users • UML Diagrams Process view: • dynamic aspects • processes and how they communicate Physical view: • topology of software components on the physical layer • network connections Development view: • programmer's perspective • software management 14
15. INNOQ / Application Container Design mit Kubernetes Kubernetes Revolution? 15
16. INNOQ / Application Container Design mit Kubernetes Resource Abstraction & Container Isolation machine 3 machine 3 Ressourcen Abstraction machine 1 machine 2 machine 3 Application A Application B Application C Service Endpoints Group of Container Configurations Network … YAML / JSON / HTTP 16
17. INNOQ / Application Container Design mit Kubernetes Deployment Replica Set POD Resource Abstraction C C C C C C C C C C C C Ing Deployment Replica Set POD Kubernetes Flat Container level management C C C Srv Srv 17
18. INNOQ / Application Container Design mit Kubernetes The Pod Ressource? Pod Container Container Container “Logical Host” • containers are tightly coupled • co-located and co-scheduled • run in a shared context • shared storage (volumes) • has its own IP 18
19. INNOQ / Application Container Design mit Kubernetes Are there Pattern for Container? 19
20. INNOQ / Application Container Design mit Kubernetes What are Container Pattern? ❏ Is valid for container as it is for OOP ❏ Provide general reusable solutions to a common problem ❏ Simplifies reuse of images ❏ Can help to modularize on container level ❏ Separation of Concerns ❏ Isolation 20
21. INNOQ / Application Container Design mit Kubernetes Single Node, Container Pattern 21
22. INNOQ / Application Container Design mit Kubernetes Pod Sidekick / Sidecar Proxy / Adapter Main App Container 1 2 3 Pod / Logical Host Container Communication Link 22
23. INNOQ / Application Container Design mit Kubernetes Sidekick / Sidecar Can be used to abstract communication or infrastructure details • Protocol switch • Authorization • Encryption • Circuit Breaker Pattern • Traffic limiting • . . . 23
24. INNOQ / Application Container Design mit Kubernetes Pod Adapter Pattern Monitoring Adapter Main App Container Complex Centralized Monitoring System Simple interface 24
25. INNOQ / Application Container Design mit Kubernetes Adapter Pattern apiVersion: v1 kind: Pod metadata: name: the-application spec: containers: - name: my-server image: innoq/web-server:latest - name: mon-adapter image: innoq/monitoring-adapter:v1.0.3 • separation of concerns • testability • reuse • flexibility dependency injection 25
26. INNOQ / Application Container Design mit Kubernetes Multi Node, Container Pattern 26
27. INNOQ / Application Container Design mit Kubernetes Application Pod Application Pod Work Queue Pattern User Supplied Container Work Execution Framework Container Work Coordination Container Application Pod User Supplied Container Work Execution Framework Container Implementation 27
28. INNOQ / Application Container Design mit Kubernetes Scatter / Gather (Messaging Pattern) Root Framework Container query():response merge([]response) User Supplied Container Merge Execution Framework Container User Supplied Container Query Execution Framework Container User Supplied Container Query Execution Framework Container User Supplied Container Query Execution Framework Container 28
29. INNOQ / Application Container Design mit Kubernetes All this are Pattern, that run distributed on a cluster of nearly infinite machines 29
30. INNOQ / Application Container Design mit Kubernetes Modularize in Container by... ➔ Architecture / Business Logic ➔ Implements NFRs at runtime ➔ Has a Framework characteristics 30
31. INNOQ / Application Container Design mit Kubernetes Further Modularization? 31
32. INNOQ / Application Container Design mit Kubernetes FaaS 32
33. INNOQ / Application Container Design mit Kubernetes Use Functions... ➔ Event trigger ➔ Business logic that can be decomposed into simple functions ➔ Simple tasks ➔ Workflow of tasks 33
34. INNOQ / Application Container Design mit Kubernetes Use Functions… advantages: ➔ No Build System ➔ No Dockerfiles ➔ No Deployment Descriptors ➔ No Server ➔ No Cluster Sizing ➔ No Workload Scaling 34
35. INNOQ / Application Container Design mit Kubernetes Kubernetes FaaS Frameworks • OpenFaaS • Kubeless • Fission • Knative • . . . 35
36. INNOQ / Application Container Design mit Kubernetes Operator Pattern 36
37. INNOQ / Application Container Design mit Kubernetes Why is is difficult to Operate things? Day 1 is quite simple (could be done by e.g. HELM) Day 2 operations: • Up- and downscaling (complex) • Backup and restore • Version bump • Resilience, Encryption, Identities etc. ➡ Operators 37
38. INNOQ / Application Container Design mit Kubernetes Operator • Operating-a-tool knowledge baked into code • Is a running Container in K8s (POD) • Talks to the K8s API as you do it with kubectl • Uses the API extensibility of K8s 38
39. INNOQ / Application Container Design mit Kubernetes etcd-X cluster Operator example ETCD etcd Operator etcd Manages Cluster CR: etcd-X Kubernetes API etcd apiVersion: etcd.database.coreos.com/v1beta2 kind: EtcdCluster metadata: name: etcd-X spec: size: 3 etcd 39
40. INNOQ / Application Container Design mit Kubernetes Operator example Kafka (Strimzi) Creates / deploys • Kafka cluster of broker nodes • ZooKeeper cluster of replicated ZooKeeper instances • Kafka Connect cluster for external data connections • Kafka MirrorMaker cluster to mirror the Kafka cluster in a secondary cluster • Kafka Exporter to extract Kafka metrics data for monitoring • Kafka Bridge to make HTTP-based requests to the Kafka cluster 40
41. INNOQ / Application Container Design mit Kubernetes Operator example Kafka (Strimzi) Manages... • Topics • Users • Connections • Security (Encryption, Authentication etc.) • . . . 41
42. INNOQ / Application Container Design mit Kubernetes Example Kafka Cluster apiVersion: kafka.strimzi.io/v1beta1 kind: Kafka metadata: name: my-cluster spec: kafka: replicas: 3 version: 0.16.1 jvmOptions: -Xmx: 8192m listeners: tls: authentication: type: tls external: type: route authentication: type: tls authorization: type: simple config: auto.create.topics.enable : "false" offsets.topic.replication.factor : 3 transaction.state.log.replication.factor : 3 transaction.state.log.min.isr : 2 storage: type: persistent-claim size: 10000Gi 42
43. INNOQ / Application Container Design mit Kubernetes Operator example vendor API complexity 43 K8s P P P P Vendor A SQL MSG Vendor B MSG Vendor C Search O
44. INNOQ / Application Container Design mit Kubernetes Usual App characteristic and Day 2 issues? • Distributed Computing • Shared Nothing Architecture • Network between Services • Encryption • Identity • Resiliency • Observability • Releasing Things 44
45. INNOQ / Application Container Design mit Kubernetes Service Mesh A Service Mesh is a decentralized infrastructure level in which functions for observability, routing, resilience and secure communication between microservices are implemented. 45
46. INNOQ / Application Container Design mit Kubernetes Basic Pattern Application Pod Proxy Main App Container Application Pod Proxy Main App Container Control Plane Tracing Monitoring 46
47. INNOQ / Application Container Design mit Kubernetes Service Mesh Features traffic management and shifting ❏ separates traffic from replicas ❏ fault injection ❏ rate limiting security ❏ mTLS between services ❏ service identity policies visualization 47
48. INNOQ / Application Container Design mit Kubernetes Options for Kubernetes Istio ❏ most buzz around ❏ huge Linkerd 2 ❏ smarter (not so feature rich as istio) stock solution with ❏ Cilium ❏ nginx ingress 48
49. INNOQ / Application Container Design mit Kubernetes Recap 49
50. INNOQ / Application Container Design mit Kubernetes Recap Container Pattern • Increases reuse, flexibility and testability • Better resource usage • YAMLs can wire functionality together Operators • Complex operation made simpler and transferable • Tool vendors regain control • Think of K8s as cloud operating system 50
51. INNOQ / Application Container Design mit Kubernetes Recap Service Mesh • Use of dynamically configurable proxies with all its associated possibilities • Security and missing-feature evaluation will eventually lead to a service mesh • Is a monster 51
52. INNOQ / Application Container Design mit Kubernetes thx 52