Deploying Distributed Databases and In-Memory Computing Platforms with Kubernetes
•
0 likes•382 views
Presented at OpenStack Summit, Berlin 2018. In this presentation, attendees will learn how Kubernetes can orchestrate a distributed database like Apache Ignite, in particular: * Cluster Assembling - database nodes auto-discovery in Kubernetes. * Database Resilience - automated horizontal scalability. * Database Availability - what’s the role of Kubernetes and the database. * Utilizing both RAM and disk - set up Apache Ignite in a way to get in-memory performance with the durability of disk.
Recommended
More Related Content
What's hot
What's hot (20)
Similar to Deploying Distributed Databases and In-Memory Computing Platforms with Kubernetes
Similar to Deploying Distributed Databases and In-Memory Computing Platforms with Kubernetes (20)
Recently uploaded
Recently uploaded (20)
Deploying Distributed Databases and In-Memory Computing Platforms with Kubernetes
- 1. © 2018 GridGain Systems, Inc. GridGain Company Confidential Deploying Distributed Databases and In-Memory Computing Platforms with Kubernetes Stephen Darlington Senior Consultant GridGain Systems
- 2. © 2018 GridGain Systems, Inc. GridGain Company Confidential Agenda Deployment with Kubernetes Memory-Only Deployments Stateful Deployments Management and Monitoring Demo Q&A
- 3. © 2018 GridGain Systems, Inc. GridGain Company Confidential Kubernetes is… …an open-source system for automating deployment, scaling, and management of containerized applications.
- 4. © 2018 GridGain Systems, Inc. GridGain Company Confidential Apache Ignite is… …a memory-centric distributed database, caching, and processing platform for transactional, analytical, and streaming workloads delivering in- memory speeds at petabyte scale.
- 5. © 2018 GridGain Systems, Inc. GridGain Company Confidential Memory-Centric Storage Scale to 1000s of Nodes & Store TBs of Data Ignite Native Persistence (Flash, SSD, Intel 3D XPoint) Third-Party Persistence Keep Your Own DB (RDBMS, HDFS, NoSQL) SQL Transactions Compute Services MLStreamingKey/Value IoTFinancial Services Pharma & Healthcare E-CommerceTravel & Logistics Telco Apache Ignite Overview
- 6. © 2018 GridGain Systems, Inc. GridGain Company Confidential Our challenges Server or VM JVM Node Server or VM JVM Node Server or VM JVM Node Server or VM JVM Node
- 7. © 2018 GridGain Systems, Inc. GridGain Company Confidential Deployment with Kubernetes
- 8. © 2018 GridGain Systems, Inc. GridGain Company Confidential Distributed Database Details • Database is a set of pods – IPs are assigned dynamically – Auto-discovery is needed • Applications Deployment – Within Kubernetes? – Not managed by Kubernetes • Stateless or Stateful?
- 9. © 2018 GridGain Systems, Inc. GridGain Company Confidential Database Pods Auto-Discovery • Kubernetes Lookup Service – Tracks a list of all Ignite pods – Gateway for remote apps • Kubernetes IP Finder – Consumes IPs from the service – Let’s node to join the cluster
- 10. © 2018 GridGain Systems, Inc. GridGain Company Confidential Service Startup
- 11. © 2018 GridGain Systems, Inc. GridGain Company Confidential IP Finder Configuration
- 12. © 2018 GridGain Systems, Inc. GridGain Company Confidential Shared Configuration
- 13. © 2018 GridGain Systems, Inc. GridGain Company Confidential Ignite Memory Usage Modes Mode Description Major Advantage In-Memory Pure In-Memory Storage Maximum perfomance possible (data is never written to disk) In-Memory + 3rd Party DB Caching layer (aka. in-memory data grid) for existing databases – RDBMS, NoSQL, etc Horizontal scalability Faster reads and writes In-Memory + Full Copy on Disk The whole data set is stored both in memory and on disk Survives cluster failures 100% on Disk + In-Memory Cache 100% of data is in Ignite native persistence and a subset is in memory Unlimited data scale beyond RAM capacity
- 14. © 2018 GridGain Systems, Inc. GridGain Company Confidential Ignite as Kubernetes Deployment Entity
- 15. © 2018 GridGain Systems, Inc. GridGain Company Confidential Stateful Deployments • Durability With StatefulSet – Data persistence to disk – Ordered restarts • Separate Volumes for – Data and indexes – WAL (aka Transaction Log) – Snapshots and backups
- 16. © 2018 GridGain Systems, Inc. GridGain Company Confidential Cluster Activation • Manual Activation on First Start • Automatic Activation on Restarts – Baseline topology usage
- 17. © 2018 GridGain Systems, Inc. GridGain Company Confidential Management and Monitoring
- 18. © 2018 GridGain Systems, Inc. GridGain Company Confidential K8 Dashboard and Ignite Web Console • Kubernetes Dashboard – For Kubernetes environment • Ignite Web Console – For Ignite cluster – Deploy Web Agent in K8
- 19. © 2018 GridGain Systems, Inc. GridGain Company Confidential Demo
- 20. © 2018 GridGain Systems, Inc. GridGain Company Confidential More information • ignite.apache.org • apacheignite.readme.io/docs/kubernetes- deployment • github.com/apache/ignite – modules/kubernetes
- 21. © 2018 GridGain Systems, Inc. GridGain Company Confidential Thank you for joining us. Follow the conversation. https://ignite.apache.org https://gridgain.com/ Any questions? @sdarlington #apacheignite @gridgain #gridgain
Editor's Notes
- Thank you open stack Thanks for coming Good to be in berlin… museum of technology, Konrad Zeus, Z1 programmable computer
- The buzzfeed version: the three tricks you need to install a compute/data grid using Kubernetes… the third will amaze you! But first, let’s define some terms.
- You know what kubernetes is… but what about in-memory compute platforms? We’re going to be talking about Ignite, but there are others that are similar
- Let’s dig down into what that really means and what it means for Kubernetes
- Lots of nodes That all need to be able to find each other Share configuration Non-shared storage (even memory only config not all nodes are fungible, but for persistence for logs, for persistent data) Dynamically add/remove nodes – without affecting the data!
- Traditionally you’d have one or more ”compute” pods (MySQL) plus some fixed disk That would kinda work for Ignite but missing the point Also, scale horizontally… all nodes equal (no master) but in charge of different data… so not equal! But no master means they need to be able to find each other
- Kubernetes already knows about nodes (pods) All we need is a little glue to link them together
- apiVersion: v1 kind: Service metadata: name: ignite namespace: ignite spec: type: LoadBalancer ports: - name: rest port: 8080 targetPort: 8080 - name: sql port: 10800 targetPort: 10800 - name: thinclients port: 10900 targetPort: 10900 selector: app: ignite
- <bean class="org.apache.ignite.configuration.IgniteConfiguration"> <property name="discoverySpi"> <bean class="org.apache.ignite.spi.discovery.tcp.TcpDiscoverySpi"> <property name="ipFinder"> <bean class=” org.apache.ignite.spi.discovery.tcp.ipfinder.kubernetes.TcpDiscoveryKubernetesIpFinder"> <!-- Assumed that RBAC is configured for `ignite` namespace. --> <property name="namespace" value="ignite"/> </bean> </property> </bean> </property> </bean>
- kubectl create configmap ignite-config --from-file=example-kube.xml
- In-Memory Use Case: in-memory caches, in-memory data grids, in-memory computations, web-session caching, real-time processing of continuous data streams. 100% on Disk + In-Memory Cache – ideal for HTAP because it doesn’t make sense to store analytical data sets in RAM or raw data for ML/DL models.
- Example here just uses a deployment, you might prefer be explicit about using a ReplicaSet Singleton instance like MySQL (persistent volume, MySQL pod, service) ReplicaSet – “should be used when your application is completely decoupled from the node” “every container is identical and replaceable” (which is true for memory-only but not for persistent) DaemonSet – “should be used when a single copy of your application must be run on all or a subset of nodes in the cluster” Dynamic Volume Provisioning (with Azure for example, using StorageClass) StatefulSet (stable in 1.9, ~ a year ago) Persistent hostname with a unique index Created and deleted in order Singleton instance like MySQL (persistent volume, MySQL pod, service) ReplicaSet – “should be used when your application is completely decoupled from the node” “every container is identical and replaceable” (which is true for memory-only but not for persistent) DaemonSet – “should be used when a single copy of your application must be run on all or a subset of nodes in the cluster” Dynamic Volume Provisioning (with Azure for example, using StorageClass) StatefulSet (stable in 1.9, ~ a year ago) Persistent hostname with a unique index Created and deleted in order apiVersion: extensions/v1beta1 kind: Deployment metadata: name: ignite-cluster spec: replicas: 2 template: metadata: labels: app: ignite spec: containers: - name: ignite-node image: apacheignite/ignite:2.6.0 env: - name: OPTION_LIBS value: ignite-kubernetes - name: CONFIG_URI value: URL_TO_CONFIG ports: ...
- Singleton instance like MySQL (persistent volume, MySQL pod, service) ReplicaSet – “should be used when your application is completely decoupled from the node” “every container is identical and replaceable” (which is true for memory-only but not for persistent) DaemonSet – “should be used when a single copy of your application must be run on all or a subset of nodes in the cluster” Dynamic Volume Provisioning (with Azure for example, using StorageClass) StatefulSet (stable in 1.9, ~ a year ago) Persistent hostname with a unique index Created and deleted in order
- Singleton instance like MySQL (persistent volume, MySQL pod, service) ReplicaSet – “should be used when your application is completely decoupled from the node” “every container is identical and replaceable” (which is true for memory-only but not for persistent) DaemonSet – “should be used when a single copy of your application must be run on all or a subset of nodes in the cluster” Dynamic Volume Provisioning (with Azure for example, using StorageClass) StatefulSet (stable in 1.9, ~ a year ago) Persistent hostname with a unique index Created and deleted in order apiVersion: apps/v1 kind: StatefulSet metadata: name: ignite namespace: ignite spec: selector: matchLabels: app: ignite serviceName: ignite ... volumeMounts: - mountPath: "/data/ignite" name: ignite-storage volumeClaimTemplates: - metadata: name: ignite-storage spec: accessModes: [ "ReadWriteOnce" ] resources: requests: storage: 1Gi
- kubectl exec -it ignite-0 --namespace=ignite -- /bin/bash cd /opt/ignite/apache-ignite-fabric/bin/ ./control.sh --activate
- [1] http://globenewswire.com/news-release/2018/07/09/1534470/0/en/The-Apache-Software-Foundation-Announces-Annual-Report-for-2018-Fiscal-Year.html [2] https://blogs.apache.org/foundation/entry/apache-in-2017-by-the
- Try to get the demo app and slides on the web