Over the last fews weeks as customers have started to ramp up their usage of CDP cloud assets (like DataHub, Experiences, etc), I have observed many of the ways they are leveraging on-prem engineering assets (code) in the cloud.  The concept of write-once-deploy-anywhere is fundamental to a well designed data strategy.  It's NOT a sales pitch.  It's a reality for enterprises who have invested in a Modern Data Architecture.  However, unlike on-prem where storage and services are tightly coupled, CDP flips that concept on its head.  We can now launch services independently and choose only the capabilities we need for the task at hand.  For example, streaming use cases typically require NiFi, Kafka, and Spark Streaming.  Each of those services would be separate DataHub clusters and scale independently.  This article focuses on using PySpark to read from a secured Kafka instance.  To be clear, this is one way (not the only way) of using PySpark to read from Kafka.  Both (DE & SM) clusters are launched via CDP control plane.     CDP DataHub assets used in this article Data Engineering Cluster (DE) Streams Messaging (SM)   Launching DataHub DE and SM clusters are well documented here.   PreWork Secure Kafka and generate certs/truststore https://docs.cloudera.com/runtime/7.0.2/kafka-securing/topics/kafka-secure-tls.html In this article I refer to the truststore as kafka.client.truststore Get a list of all Kafka brokers and ports For this article I will call the brokers k1.cloudera.com:9093, k2.cloudera.com:9093, k3.cloudera.com:9093 Create a kafka.properties file       security.protocol=SASL_SSL sasl.mechanism=PLAIN ssl.truststore.location=/home/sunilemanjee/kafka.client.truststore.jks ssl.truststore.password=password sasl.jaas.config=org.apache.kafka.common.security.plain.PlainLoginModule required username="machine-user" password="password;       Create a Kafka jaas file called jaas.conf (this can be named whatever you like)       KafkaClient { org.apache.kafka.common.security.plain.PlainLoginModule required username="machine-user" password="password"; };         Create a Kafka topic The easiest way to create a Kafka topic is via SMM (Streamings Messaging Manager) which is shipped with Streams Messaging cluster.  Click on the SMM URL within DataHub and the click on "Topics" located on the right menu bar.   Click on "Add New" to create a new Kafka topic.   Enter the topic name "demo", set partitions to 1, and clean up policy to "delete".   The demo topic should now be available.   Generate Data For PySpark to consume from a secured instance of Kafka, we need the ability to write to Kafka.  Here we will use Kafka console.  SSH into one of the broker nodes Update k*.cloudera.com:9093 with your broker list and ports Upload kafka.properties (created early) onto this node Update the location of your kafka.properties file After the below command is executed, you can start to write data (messages) to Kafka.  We will come back to this in a moment.       kafka-console-producer --broker-list k1.cloudera.com:9093, k2.cloudera.com:9093,k3.cloudera.com:9093 --producer.config /home/c sunilemanjee/kafka.properties --topic demo         Read from Kafka using PySpark SSH into any node within the DE cluster Uploaded jaas.conf and kafka.client.truststore Update the location of jaas.conf and kafka.client.truststore Launch PySpark shell using the following command         pyspark --files "/home/csunilemanjee/jaas.conf#jaas.conf,/home/sunilemanjee/kafka.client.truststore.jks#kafka.client.truststore.jks" --driver-java-options "-Djava.security.auth.login.config=/home/sunilemanjee/jaas.conf" --conf "spark.executor.extraJavaOptions=-Djava.security.auth.login.config=/home/sunilemanjee/jaas.conf"         Once PySpark shell is up, it may be easier to store the Kafka brokers in a variable like this:       KAFKA_BROKERS = "k1.cloudera.com:9093,k2.cloudera.com:9093,k3.cloudera.com:9093"       Create a  structured stream to read from Kafka.  Update the following kafka.ssl.truststore.location kafka.ssl.truststore.password username password       df_kafka = spark.readStream.format("kafka").option("kafka.bootstrap.servers", KAFKA_BROKERS).option("subscribe", "demo").option("kafka.security.protocol", "SASL_SSL").option("kafka.sasl.mechanism", "PLAIN").option("kafka.ssl.truststore.location", "./kafka.client.truststore.jks").option("kafka.ssl.truststore.password", "password").option("kafka.sasl.jaas.config", "org.apache.kafka.common.security.plain.PlainLoginModule required username=\"machine-user\" password=\"password\"serviceName=\"kafka\";").load().selectExpr("CAST(key AS STRING)", "CAST(value AS STRING)").writeStream.format("console").trigger(continuous="1 second").start()       To start viewing Kafka message on the console (pyspark shell) from the "demo" topic       stream = df_kafka.writeStream.format("console").start() stream.awaitTermination() ##once you are finished, to kill the stream run this stream.stop()       Go back to your kafka console and start write messaging (anything you like).  You will see those messages show up in your PySpark Shell console.   That's it.  Again this is one (not the only way) to use PySpark to consume from a secured Kafka instance.  I see as an emerging pattern in the CDP for streaming use cases.  Enjoy.     ... View more

      My Kubernetes series  (Part1, Part2) was strictly focused on MiNiFi on K8S.  NiFi and MiNiFi may communicate over Site 2 Site; however, often the pattern is to leverage Kafka for a clean message handoff.  NiFi within these patterns is generally the central router and transformer of messages.  Think of it like "FedEx" for data.   Till now most have deployed NiFi on bare metal or VMs.  Natural evolution kicks in. Deploy NiFi on k8s and yes it's super simple.  In this article I will demonstrate how to deploy both a NiFi and ZooKeeper cluster (none being a single pod!) on Azure Kubernetes Service (AKS); however, all artifacts may be leveraged to deploy on virtually any kubernetes offering.   Prerequisites Some knowledge of Kubernetes and NiFi AKS / k8s Cluster kubectl cli NiFi image available in a registry (ie dockerhub)   Deployment All the contents in the ymls below can be placed into single file for deployment.  For this demonstration, chucking it into smaller components makes it easier to explain.   I have loaded a NiFi image into Azure Container Repository.  You can use the NiFi image available here in DockerHub.     ZooKeeper NiFi uses ZooKeeper for several state management functions.  More on that here.  ZooKeeper for NiFi can be deployed using embedded or stand alone mode.  Here I will deploy 3 pods of ZK on k8s.   Deploying ZK on k8s is super simple.   kubectl apply -f https://k8s.io/examples/application/zookeeper/zookeeper.yaml   After a few minutes 3 pods of ZK will be available for NiFi to use.  Once ZK pods become available, proceed  to deploy NiFi on k8s   kubectl get pods -w -l app = zk   NiFi Below is the k8s deployment yml for NiFi (named nifi.yml).  Few fields to highlight replicas This will be number of NiFi pods to deploy (Cluster) Image Replace with your image location           apiVersion: extensions/v1beta1 kind: Service #+ apiVersion: v1 #+ metadata: #+ name: nifi-service #+ spec: #+ selector: #+ app: nifi #+ ports: #+ - protocol: TCP #+ targetPort: 8080 #+ port: 8080 #+ name: ui #+ - protocol: TCP #+ targetPort: 9088 #+ port: 9088 #+ name: node-protocol-port #+ - protocol: TCP #+ targetPort: 8888 #+ port: 8888 #+ name: s2s #+ type: LoadBalancer #+ --- kind: Deployment apiVersion: extensions/v1beta1 metadata: name: nifi spec: replicas: 2 selector: matchLabels: app: nifi template: metadata: labels: app: nifi spec: hostNetwork: true dnsPolicy: ClusterFirstWithHostNet containers: - name: nifi-container image: sunmanregistry.azurecr.io/nifi:latest ports: - containerPort: 8080 name: http - containerPort: 22 name: ssh resources: requests: cpu: ".5" memory: "6Gi" limits: cpu: "1" env: - name: VERSION value: "1.9" - name: NIFI_CLUSTER_IS_NODE value: "true" - name: NIFI_CLUSTER_NODE_PROTOCOL_PORT value: "9088" - name: NIFI_ELECTION_MAX_CANDIDATES value: "1" - name: NIFI_ZK_CONNECT_STRING value: "zk-0.zk-hs.default.svc.cluster.local:2181,zk-1.zk-hs.default.svc.cluster.local:2181,zk-2.zk-hs.default.svc.cluster.local:2181"             To deploy this NiFi k8s manifest:       kubectl apply -f nifi.yml           Few minutes later you will have a fully functioning NiFi cluster on Kubernetes       I use super simple a lot.  I know.  Deploying MiNiFi & NiFi on k8s is just that, super simple.  Enjoy.     ... View more

  Part 2 of Autoscaling MiNiFi on K8S is focused on deploying the artifacts on AKS - Amazon Kubernetes Service.  My knew jerk reaction was all Kubernetes as a Service would play well but that is definitely not the case.  Hence why GCP Anthos product direction for this space is a key. The net-net of my observation is k8s app deployment on any single cloud vendor would cause deployment complexities any other k8s deployment, cloud or OnPrem..  Vendor lock in theory is ALIVE and WELL.  In Azure I leveraged ACS for EFM and NiFi Registry; however, the natural evolution was to deploy EFM and NiFi Registry (NR) on K8S.  EFM, NR, and MiNiFi are integrated components (refer to part 1 on architecture).  I will leverage several key out of the box k8s components to make this all work together.  The good news is, the deployment is super simple!   Prerequisites Some knowledge of Kubernetes and AKS AKS Cluster kubectl cli eksctl cli VPC 2 public subnets within a VPC NR, EFM, and MiNiFi images uploaded to and available ECS Refer to part 1 on image locations   Create a AKS Cluster eksctl Makes this simple.  I tried using aws eks and it was painful.       eksctl create cluster \ --name sunman-k8s \ --version 1.13 \ --nodegroup-name standard-workers \ --node-type t3.medium \ --nodes 3 \ --nodes-min 1 \ --nodes-max 4 \ --vpc-public-subnets=subnet-067d0ffbc09152382,subnet-037d8c6750c5de236 \ --node-ami auto       Deployment All the contents in the ymls below can be placed into single file for deployment.  For this demonstration, chucking it into smaller components makes it easier to explain. NiFi Registry (NR) Edge Flow Manager has a dependency on NR.  Flow versions are stored in NR.  Here is nifiregistry.yml.       apiVersion: extensions/v1beta1 kind: Deployment metadata: name: nifiregistry spec: replicas: 1 selector: matchLabels: app: nifiregistry template: metadata: labels: app: nifiregistry spec: containers: - name: nifiregistry-container image: your-image-location/nifiregistry ports: - containerPort: 18080 name: http - containerPort: 22 name: ssh resources: requests: cpu: ".5" memory: "2Gi" limits: cpu: "1" env: - name: VERSION value: "11" --- kind: Service #+ apiVersion: v1 #+ metadata: #+ name: nifiregistry-service #+ spec: #+ selector: #+ app: nifiregistry #+ ports: #+ - protocol: TCP #+ targetPort: 18080 #+ port: 80 #+ name: http #+ - protocol: TCP #+ targetPort: 22 #+ port: 22 #+ name: ssh #+ type: LoadBalancer #+ loadBalancerSourceRanges: - 0.0.0.0/0       Update the following line in nifiregistry.yml with the location of your NR image.     image: your-image-location/nifiregistry     Also take note the load balancer for NR is open to the world.  You may want to lock this down.   Deploy NR on k8s     kubectl apply -f nifiregistry.yml       Edge Flow Manager (EFM) Next deploy EFM on k8s.   Here is efm.yml     apiVersion: extensions/v1beta1 kind: Deployment metadata: name: efm spec: replicas: 1 selector: matchLabels: app: efm template: metadata: labels: app: efm spec: containers: - name: efm-container image: your-image-location/efm ports: - containerPort: 10080 name: http - containerPort: 22 name: ssh resources: requests: cpu: ".5" memory: "2Gi" limits: cpu: "1" env: - name: VERSION value: "11" - name: NIFI_REGISTRY_ENABLED value: "true" - name: NIFI_REGISTRY_BUCKETNAME value: "testbucket" - name: NIFI_REGISTRY value: "<a href="<a href="<a href="<a href="<a href="http://nifiregistry-service.default.svc.cluster.local" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a>" target="_blank"><a href="http://nifiregistry-service.default.svc.cluster.local</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a>>" target="_blank"><a href="<a href="http://nifiregistry-service.default.svc.cluster.local</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a>" target="_blank"><a href="http://nifiregistry-service.default.svc.cluster.local</a</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a</a>>>" target="_blank"><a href="<a href="<a href="http://nifiregistry-service.default.svc.cluster.local</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a>" target="_blank"><a href="http://nifiregistry-service.default.svc.cluster.local</a</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a</a>>" target="_blank"><a href="<a href="http://nifiregistry-service.default.svc.cluster.local</a</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a</a>" target="_blank"><a href="http://nifiregistry-service.default.svc.cluster.local</a</a</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a</a</a>>>>" target="_blank"><a href="<a href="<a href="<a href="http://nifiregistry-service.default.svc.cluster.local</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a>" target="_blank"><a href="http://nifiregistry-service.default.svc.cluster.local</a</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a</a>>" target="_blank"><a href="<a href="http://nifiregistry-service.default.svc.cluster.local</a</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a</a>" target="_blank"><a href="http://nifiregistry-service.default.svc.cluster.local</a</a</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a</a</a>>>" target="_blank"><a href="<a href="<a href="http://nifiregistry-service.default.svc.cluster.local</a</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a</a>" target="_blank"><a href="http://nifiregistry-service.default.svc.cluster.local</a</a</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a</a</a>>" target="_blank"><a href="<a href="http://nifiregistry-service.default.svc.cluster.local</a</a</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a</a</a>" target="_blank"><a href="http://nifiregistry-service.default.svc.cluster.local</a</a</a</a" target="_blank">http://nifiregistry-service.default.svc.cluster.local</a</a</a</a</a>>>>>" --- kind: Service #+ apiVersion: v1 #+ metadata: #+ name: efm-service #+ spec: #+ selector: #+ app: efm #+ ports: #+ - protocol: TCP #+ targetPort: 10080 #+ port: 80 #+ name: http #+ - protocol: TCP #+ targetPort: 22 #+ port: 22 #+ name: ssh #+ type: LoadBalancer #+ loadBalancerSourceRanges: - 0.0.0.0/0       Update the following line in efm..yml with the location of your EFM image.     image: your-image-location/efm     Also take note the load balancer for EFM is open to the world.  You may want to lock this down.   Deploy EFM on k8s     kubectl apply -f efm.yml         MiNiFi Lastly,  deploy MiNiF on k8s.   Here is minifi.yml     apiVersion: extensions/v1beta1 kind: Deployment metadata: name: minifi spec: replicas: 1 selector: matchLabels: app: minifi template: metadata: labels: app: minifi spec: containers: - name: minifi-container image: your-image-location/minifi-azure-aws ports: - containerPort: 10080 name: http - containerPort: 6065 name: listenhttp - containerPort: 22 name: ssh resources: requests: cpu: "500m" memory: "1Gi" limits: cpu: "1" env: - name: NIFI_C2_ENABLE value: "true" - name: MINIFI_AGENT_CLASS value: "listenSysLog" - name: NIFI_C2_REST_URL value: "<a href="<a href="<a href="<a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a>" target="_blank"><a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a</a>>" target="_blank"><a href="<a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a</a>" target="_blank"><a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a</a</a>>>" target="_blank"><a href="<a href="<a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a</a>" target="_blank"><a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a</a</a>>" target="_blank"><a href="<a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a</a</a>" target="_blank"><a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a</a</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/heartbeat</a</a</a</a>>>>" - name: NIFI_C2_REST_URL_ACK value: "<a href="<a href="<a href="<a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a>" target="_blank"><a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a</a>>" target="_blank"><a href="<a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a</a>" target="_blank"><a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a</a</a>>>" target="_blank"><a href="<a href="<a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a</a>" target="_blank"><a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a</a</a>>" target="_blank"><a href="<a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a</a</a>" target="_blank"><a href="http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a</a</a" target="_blank">http://efm-service.default.svc.cluster.local/efm/api/c2-protocol/acknowledge</a</a</a</a>>>>" --- kind: Service #+ apiVersion: v1 #+ metadata: #+ name: minifi-service #+ spec: #+ selector: #+ app: minifi #+ ports: #+ - protocol: TCP #+ targetPort: 10080 #+ port: 10080 #+ name: http #+ - protocol: TCP #+ targetPort: 9877 #+ port: 9877 #+ name: tcpsyslog - protocol: TCP #+ targetPort: 9878 #+ port: 9878 #+ name: udpsyslog - protocol: TCP #+ targetPort: 22 #+ port: 22 #+ name: ssh #+ - protocol: TCP #+ targetPort: 6065 #+ port: 6065 #+ name: listenhttp #+ type: LoadBalancer #+ loadBalancerSourceRanges: - 0.0.0.0/0         Update the following line in minifi..yml with the location of your MiNiFI image.     image: your-image-location/minifi-azure-aws     Also take note the load balancer for MiNiFi is open to the world.  You may want to lock this down.   Deploy MiNiFi on k8s     kubectl apply -f minifi.yml         Thats it!  Part 1 of this series demonstrated how to autoscale MiNiFi and those same k8s commands can be used here to scale this out properly.  The next part of this series will add the concept of k8s stateful sets and their impact EFM/NR/MiNiFi for a resilient backend persistence layer.   ... View more

MiNiFi (Java Version) is essentially NiFi with a few differences and hence why it runs so darn well on containers/Kubernetes. The use case is to have a single management console (Edge Flow Manager) to manage 0 + many MiNiFi agents which require autoscaling on Kubernetes based on some arbitrary metrics...for example CPU/RAM threshold. EFM and NiFi Registry are required but don't need autoscaling; therefore, these services will be deployed on Azure Container Service. MiNiFi on the other hand often benefits from autoscaling and hence it will be deployed on Azure Kubernetes Service. Required for this demonstration Azure subscription Container Registry Demo will leverage Azure Container Registry Kubernetes Service Demo will leverage Azure Kubernetes Service Azure CLI The following images need to be stored in Azure Container Registry Edge Flow Manager https://github.com/sunileman/efm1.0.0.0-docker NiFi Registry https://github.com/sunileman/NiFi-Registry-Service MiNiFi (Java) https://github.com/sunileman/CEM1.0-Java-MiNiFi This image will come precooked with Azure/AWS NARs Architecture This is a 10k foot view of the architecture. EFM communicates with MiNiFi agents about the work they need to do. EFM also communicates with NiFi Registry to store/version control flows will get passed to the MiNiFi agents. Deploy NiFi Registry and EFM on Azure Container Service Since EFM and Registry don't really benefit from autoscaling, they both are great fit for Azure container service (Mostly Static installs). ACS will guarantee EFM and NiFi registry are alway up with 1 container instance each. EFM, MiNiFi, and Registry have all been imported into my container registry on azure. Create NiFi Registry on ACS NiFi Registry variables to note --name Name of the nifi registry container --dns-name-label Prefix for the dns on the registry service. This will be used as an input into EFM container environment variable az container create --resource-group sunmanCentralRG --name mynifiregistry --image sunmanregistry.azurecr.io/nifiregistry:latest --dns-name-label nifiregistry --ports 18080 --registry-username ****** --registry-password ****** Create EFM on ACS EFM variables to note --NIFI_REGISTRY should match NiFi registry Container DNS (fully qualified server name) --dns-name-label DNS prefix az container create --resource-group sunmanCentralRG --name efm --image sunmanregistry.azurecr.io/efm:latest --dns-name-label myefm --ports 10080 --registry-username ***** --registry-password **** --environment-variables 'NIFI_REGISTRY_ENABLED'='true' 'NIFI_REGISTRY_BUCKETNAME'='testbucket' 'NIFI_REGISTRY'='http://mynifiregistry.centralus.azurecontainer.io:18080' Create a 'testbucket' on NiFi Registry MiNiFi flows will be designed using EFM and stored in the NiFi Registry bucket 'testbucket'. This bucket name was identified as a variable during EFM container was creation. 'NIFI_REGISTRY_BUCKETNAME'='testbucket' NiFi registry will be available under YourNiFiRegistryDSN:18080/nifi-registry/ . For example http://mynifiregistry=y.centralus.azurecontainer.io:18080/nifi-registry/ Click on "NEW BUCKET", Enter bucket name - testbucket Click create Validate EFM is up EFM UI will be available under http://YourEfmDnsPrefix.centralus.azurecontainer.io:10080/efm/ui for example http://myefm.centralus.azurecontainer.io:10080/efm/ui Run MiNiFi Kubernetes Deployment The easiest way to run a deployment in k8s is to build a manifest file. To learn more about k8s manifest files here. Look for < > in the manifest below, as these are the variables a change prior to your deployment (only a few, super simple). Variable to Note MINIFI_AGENT_CLASS This will be the agent class published to EFM. To learn more about EFM, go here Kubernet Manifest File: apiVersion: extensions/v1beta1 kind: Deployment metadata:   name: minifi spec:   replicas: 1   selector:     matchLabels:       app: minifi   template:     metadata:       labels:         app: minifi     spec:       containers:       - name: minifi-container         image: <Your Containe Registry>/minifi-azure-aws:latest         ports:         - containerPort: 10080           name: http         - containerPort: 6065           name: listenhttp           - containerPort: 22           name: ssh         resources:           requests:             cpu: ".05"             memory: "1Gi"           limits:             cpu: "1"         env:         - name: NIFI_C2_ENABLE           value: "true"         - name: MINIFI_AGENT_CLASS           value: "test"         - name: NIFI_C2_REST_URL           value: "http://<Your EFM servername>.centralus.azurecontainer.io:10080/efm/api/c2-protocol/heartbeat"         - name: NIFI_C2_REST_URL_ACK           value: "http://<Your EFM servername>.centralus.azurecontainer.io:10080/efm/api/c2-protocol/acknowledge" --- kind: Service             #+ apiVersion: v1            #+ metadata:                 #+   name: minifi-service     #+ spec:                     #+   selector:               #+     app: minifi            #+   ports:                  #+   - protocol: TCP         #+     targetPort: 10080     #+     port: 10080              #+     name: http            #+   - protocol: TCP         #+     targetPort: 22        #+     port: 22              #+     name: ssh             #+   - protocol: TCP         #+     targetPort: 6065        #+     port: 6065              #+     name: listenhttp             #+   type: LoadBalancer      #+ Once the manifest file has been updated, store it as minifi.yml (this can be any name). Deploy on k8s using kubectl apply -f minifi.yml output sunile.manjee@hwx:~/Documents/GitHub/AKS-YAMLS(master⚡) » kubectl apply -f minifi.yml deployment.extensions/minifi created service/minifi-service created sunile.manjee@hwx:~/Documents/GitHub/AKS-YAMLS(master⚡) » MiNiFi has been successfully deployed. To verify successful deployment visit EFM. EFM should show the agent class name 'test' matching the class name used in the minifi k8s manifest file. Open the class and design any flow. Here I simply used GenerateFlowFile and terminated success relationship with 3 concurrent threads Click on publish and soon thereafter MiNiFi will be executing the flow. AutoScale MiNiFi At this time a single MiNiFi container/agent is executing flows. I purposefully set MiNiFi CPU allocation (manifest file) to a small number to force the autoscaling. First lets check the number of minifi pods running on k8s Single MiNiFi pod. Lets check if autoscaling is enabled for this deployment To enable autoscaling on k8s: kubectl autoscale deployment minifi --cpu-percent=25 --min=1 --max=3 minifi is the deployment name. If CPU utilization exceeds 25%, the autoscaler increases the pods up to a maximum of 3 instances. A minimum of 1 instances is then defined for the deployment Verify autoscaling is enabled on the minifi deployment Number of minifi pods after autoscaling was enabled (3). Kubernetes added 2 additional MiNiFi pods. Lets kill one of the pods and see what happens Kubernetes immediately launched a new MiNiFi container after a MiNiFi pod was killed. Enjoy AutoScale on MiNiFi! ... View more

Building an Apache NiFi processor is super easy. I have seen/read several articles on how to get started by executing maven commands via CLI. This article is geared towards individuals who like to use an IDE (specially IntelliJ) to do the imports instead of running via CLI. On IntelliJ click on create project, Check "Create from archetype", click on "ADD Archetype" and enter the following GroupId: org.apache.nifi ArtifactId: nifi-processor-bundle-archetype Version: <YourVersionOfNifi> and then click "OK" Now your new NiFi Archetype has been created. Select it Enter GroupId, ArtifactId, and Version of your choice A final attribute we need to add is artifactBaseName. This is mandatory. Click on "+" and enter Name: artifactBaseName Value: whatEverYouLike Now a project ready to build a custom processor. Enjoy! ... View more

I came across an article on how to setup NiFi to write into ADLS which required cobbling together various integration pieces and launching HDI. Since then there have been many updates in NiFi enabling a much easier integration. Combined with CloudBreak's rapid deployment of a HDF clusters provides an incredible ease of user experience. ADLS is Azure's native cloud storage (Look and feel of HDFS) and the capability to read/write via NiFi is key. This article will demonstrate how use use a CloudBreak Recipe to rapidly deploy a HDF NiFI "ADLS Enabled" cluster. Assumptions A CloudBreak instance is available Azure Credentials available Moderate familiarity with Azure Using HDF 3.2+ From Azure you will need: ADLS url Application ID Application Password Directory ID NiFi requires ADLS jars, core-site.xml, and hdfs-site.xml. The recipe I built will fetch these resources for you. Simply download the recipe/script from: https://s3-us-west-2.amazonaws.com/sunileman1/scripts/setAdlsEnv.sh Open it and scroll all the way to the bottom Update the following: Your_ADLS_URL: with your adls url Your_APP_ID: with your application ID Your_APP_Password: with your application password Your_Directory_ID: with your directory id Once the updates are completed, simply add the script under CloudBreak Recipes. Make sure to select "post-cluster-install" Begin provisioning a HDF cluster via CloudBreak. Once the Recipes page is shown, add the recipe to run on the NiFi nodes. Once cluster is up use the PutHDFS processor to write to ADLS. Configure PutHDFS Properties Hadoop Configuration Resources: /home/nifi/sities/core-site.xml,/home/nifi/sites/hdfs-sites.xml Additional Classpath Resources: /home/nifi/adlsjars Directory: / The above resources are all available on each node due to the recipe. All you have to do is call the location of the resources in the PutHDFS processor. That's it! Enjoy ... View more

I came across an article on how to setup NiFi to write into ADLS which required users to cobble together various integration pieces and launching HDI. Since then there have been many updates in NiFi enabling a much easier integration. Combine with CloudBreak's rapid deployment of a HDF cluster provides incredible ease of use. ADLS is native cloud storage provided by Azure (Look and feel of HDFS) and the capabilities to read/write via NiFi is key. This article will demonstrate how use use CloudBreak to rapidly deploy a HDF NiFI "ADLS Enabled" cluster. ... View more

Objective of this article is to demonstrate how to rapidly deploy a demo Druid & LLAP cluster preloaded with 20 years (nearly 113 million records) of airline data ready for analytics using CloudBreak on any IaaS. Entire deployment is UI driven without the need for a large overhead of administration. All artifacts mentioned in this article are publicly available for reuse to try on your own Prolegomenon Time series is an incredible capability highly leveraged within the IoT space. Current solution sets offer non scalable & expensive or distributed processing engines lacking low latency OLAP speeds. Druid is an OLAP time series engine backed by a Lambda architecture. Druid out of the box SQL capabilities are severely limited and without join support. Layering HiveQL over Druid brings the best of both worlds. Hive 3 also offers HiveQL over Kafka essentially making Hive a true SQL federation engine. With Druid’s native integration with Kafka, streaming data from Kafka directly into Druid while executing real time SQL queries via HiveQL offers a comprehensive Time Series solution for the IoT space. On with the demo.... To begin the demonstration, launch a CloudBreak deployer instance on any IaaS or on prem VM. Quick start makes this super simple. Launching CloudBreak deployer is well documented here. Once the CloudBreak deployer is up, add your Azure, AWS, GCP, or OpenStack credentials within the CloudBreak UI. This will allow deployment of the same cluster on any IaaS. Druid Blue Print To launch a Druid/LLAP cluster, an Ambari blue print will be required. Click On Blueprints Click on CREATE BLUEPRINT Name the blue print and enter the following url to import it into CloudBreak https://s3-us-west-2.amazonaws.com/sunileman1/ambari-blueprints/hdp3/druid+llap+Ambari+Blueprint Recipes Druid Requires a MetaStore. Import the recipe to create the MetaStore into CloudBreak Under Cluster Extensions, click on Recipes Enter a name for the recipe and select "pre-ambari-start" to run this recipe prior to Ambari starting Under URL enter the following to import the recipe into CloudBreak https://s3-us-west-2.amazonaws.com/sunileman1/scripts/druid+metastore+install.sh This cluster will come preloaded with 20 years of airline data Enter a name for recipe and select "post-cluster-install to run this recipe once HPD services are up Under URL enter the following to import the recipe into CloudBreak https://s3-us-west-2.amazonaws.com/sunileman1/scripts/airline-data.sh Create a Cluster Now that all recipes are in place, next step is to create a cluster Select a IaaS to deploy on (credential) Enter Cluster Name Select HDP 3.0 Select cluster type: Druid LLAP HPD 3 This is the new blue print which was imported in previous steps Select an image. Base image will do for most deployments Select instance types Note - I used 64 GB of ram per node. Additionally, I added 3 compute nodes. Select network to deploy the cluster on. If one is not pre-created, CloudBreak will create one CloudBreak can be configured to use S3, ADLS, WASB, GCS Configuring CloudBreak for S3 here Configuring CloudBreak for ADLS here Configuring CloudBreak for WASB here Configuring CloudBreak for GCS On the Worker node attach get-airline-data recipe On the Druid Broker node attach druid-metastore-install External MetaStores (databases) can be bootstrapped to the cluster. This demo does not require it. Knox will not be used for this demo Attach a security group to each host group. If a SG is not pre-created, CloudBreak will create one Lastly, provide an Ambari password and ssh key Cluster deployment and provision will begun. Within the next few minutes a cluster will be ready and an Ambari URL will be available. Zeppelin NoteBook can be imported using the url below. https://s3-us-west-2.amazonaws.com/sunileman1/Zeppelin-NoteBooks/airline_druid.json Here I demonstrated how to rapidly launch a Druid/LLAP cluster preloaded with airline data using CloudBreak. Enjoy Druid, it's crazy fast. HiveQL makes Druid easy to work with. CloudBreak makes the deployment super quick. ... View more

Is it possible to rebuild druid segment from hive? I am reading druid documentation and don't see an easy way to rebuild a druid segment from hive. Are there any solutions or patterns to rebuild druid segments within hive? I want to avoid direct druid api if at all possible. ... View more

On HDP 3.0.1 I have druid backed by derby and hive fails it could not load mysql jdbc driver. Create hive table back by druid: CREATE TABLE airline_druid STORED BY 'org.apache.hadoop.hive.druid.DruidStorageHandler' TBLPROPERTIES ( "druid.segment.granularity" = "MONTH", "druid.query.granularity" = "DAY") AS SELECT cast(Year || '-' || Month || '-' || DayofMonth as timestamp) as `__time`, cast(Year as string) Year, cast(Month as string) Month, cast(DayofMonth as string) DayofMonth, cast(UniqueCarrier as string) UniqueCarrier, cast(Origin as string) Origin, cast(Dest as string) Dest, cast(CancellationCode as string) CancellationCode, Distance, Cancelled FROM airline_raw limit 100 I get the following error: ERROR : FAILED: Execution Error, return code 1 from org.apache.hadoop.hive.ql.exec.DDLTask. org.skife.jdbi.v2.exceptions.UnableToObtainConnectionException: java.sql.SQLException: Cannot load JDBC driver class 'com.mysql.jdbc.Driver' INFO : Completed executing command(queryId=hive_20181108214414_80b7f4e8-3b79-4623-bb6b-293f5c7fe2b2); Time taken: 241.156 seconds Error: Error while processing statement: FAILED: Execution Error, return code 1 from org.apache.hadoop.hive.ql.exec.DDLTask. org.skife.jdbi.v2.exceptions.UnableToObtainConnectionException: java.sql.SQLException: Cannot load JDBC driver class 'com.mysql.jdbc.Driver' (state=08S01,code=1) I did find this HCC Post: https://community.hortonworks.com/content/supportkb/155363/error-failed-execution-error-return-code-1-from-ca.html What is weird is the error has asking for mysql driver. The druid metadata is backed by derby. In fact I did set those params set hive.druid.metadata.username=druid; set hive.druid.metadata.password=*****; set hive.druid.metadata.uri=jdbc:derby://localhost:1527/druid any ideas? ... View more

This article will demonstrate how to rapidly launch a Spark cluster on AWS via CloudBreak. The prerequisites are documented here. Once you have a AWS account and credentials, launching a Spark cluster is simple. CloudBreak is your command and control center UI for rapidly launching clusters on AWS, Az\ure, GCP, and on prem. Once the UI is up, add your AWS credentials Select AWS as your cloud provider Select the method for authentication. Key or Role. I prefer role but both work well. Click on the help button and follow the directions on how to setup auth for either method. Now that credentials have been setup, cluster creation may begin. Click on "Clusters" on top left and then click on "Create Cluster" on top right Select Advanced on top left Select Credential: Your AWS Credentials Cluster Name: Name your cluster Region: AWS Region Platform Version: HDP 3.0 Cluster Type: To run data science and ETL workloads, select HDP 3.0 Data Science blueprint Click Next Choose Image Type: Select Base Image Choose Image: Select Redhat from drop down list Here options are presented to select AWS instance types. If doing this for the first time, the defaults are fine. Click Next Select the VPC this cluster will be deployed to. If a VPC has not been pre-created, CloudBreak will create one. Click Next Clusters launched on AWS can access data stored in s3. Instructions on enabling s3 access is here. Recipes are actions performed on nodes before and/or after cluster install. If custom actions are not required, click next Next option is to configure auth and metadata database. For those just beginning, click next. Knox is highly recommended; however, if running for first time then disable it. Select AWS security group (SG). If SG has not been pre-created CloudBreak will create one. Lastly, enter a password for the admin user and ssh key. SSH key will be required if there is interest in ssh'ing into the nodes. The cluster may take 5-15 minutes to deploy. Once the cluster is up the Ambari URL will be available. Enjoy! ... View more

I am getting the following error on zeppelin: %pyspark df_single_recipes = df[df.Lotdf_nrecipes[df_nrecipes == 1].index Traceback (most recent call last): File "/tmp/zeppelin_pyspark-3154198105554295939.py", line 364, in <module> code = compile('\n'.join(stmts), '<stdin>', 'exec', ast.PyCF_ONLY_AST, 1) File "<stdin>", line 2 __zeppelin__._displayhook() ^ SyntaxError: invalid syntax Any ideas how to fix this? ... View more

I am using cloudbreak to launch simple instance of spark and hbase for HDP 3.0. Spark and hbase are install and running; however, phoenix query server is not available or was not installed. I create a custom blueprint and have phoenix query server in the blueprint. Any ideas what I am doing wrong? The blueprint is here: https://github.com/sunileman/ambari-blueprints/blob/master/spark-hbase I am able to manually install PQS via ambari host page. However, why is PQS not installed automtically when I have it called out in the blueprint. ... View more

I am trying to find an example of running tensorflow on yarn using GPUs. I have read the blog here: https://hortonworks.com/blog/distributed-tensorflow-assembly-hadoop-yarn/ But the tensorflow example provided does not run on gpu. Feedback appreciated. ... View more

Log Forwarding/Ingestion Patterns Log forwarding & ingestion is a key starting point for many logging initiatives such as log analytics, cyber security, anomaly & bot detection, etc etc. This article will focus few (not comprehensive) patterns for log forwarding/ingestion using NiFi. Commonly rsyslog is used to capture and ship log messages.“Rsyslog is an open-source software utility used on UNIX and Unix-like computer systems for forwarding log messages in an IPnetwork. It implements the basic syslog protocol, extends it with content-based filtering, rich filtering capabilities, flexible configuration options and adds features such as using TCP for transport.” More on how to configure rsyslog: here NiFi is able to ingest messages from rsyslog over TCP or UDP via ListenSysLog processor. This allows for little to no coding. Patterns Pattern A A minimalist design. Rsyslog is configured to simply forward log messages to a NiFi cluster. Rsyslog /etc/rsyslog.conf file needs to be configured to forward messages to a NiFi port identified in ListenSysLog processor. Pattern B A MiNiFi listen socket design. MiNiFi is installed on the server(s) leveraging ListenSysLog processor. This pattern offers end to end data linage along with more rich operational capabilities compared to Pattern A. MiNiFi via ListenSysLog will capture rsyslog messages and ship them to NiFi via S2S (site 2 site). Rsyslog is configured to simply forward log messages to a locally installed MiNiFi instance (localhost:port). Rsyslog /etc/rsyslog.conf file needs to be configured to forward messages to a the local MiNiFi port identified in ListenSysLog processor. This design will provide at least once message delivery guarantee. Pattern C A MiNiFi tail file design. MiNiFi is installed on the server(s) leveraging TailFile processor unlike Pattern B using ListenSyslog. Both pattern A and B offer end to end data linage and rich operational capabilities. MiNiFi will capture log messages by tailing a directory of files or a file and ship them to NiFi via S2S (site 2 site). Identify a log file to tail (ie /var/log/messages) or a directory for files, start MiNiFi and the log messages will start flow from the server(s) to NiFi. This design will provide at least once message delivery guarantee. These are a few but common pattens I have developed & implemented in the field with success. Happy log capturing! ... View more

During launch of HDP or HDF on azure via cloudbreak, if the following provisioning error is thrown (Check cloudbreak logs): log:55 INFO  c.m.a.m.r.Deployments checkExistence - [owner:xxxxx] [type:STACK] [id:2] [name:sparky] [flow:xxx] [tracking:] <-- 404 Not Found https://management.azure.com/subscriptions/xxxxxx/resourcegroups/spark. (104 ms, 92-byte body)/cbreak_cloudbreak_1 | 2018-09-05 14:25:22,882 [reactorDispatcher-24] launch:136 ERROR c.s.c.c.a.AzureResourceConnector - [owner:xxxxxx] [type:STACK] [id:2] [name:sparky] [flow:xxxxxx] [tracking:] Provisioning error: This means the instance type selected is not available within the region. Please change region where instance is available or change to instance type which is available within region. ... View more