images/blog/202107_ecomp/accumulo-compactor-muchos.yaml

# The following instance properties are mapped into the containers as
# accumulo.properties and will be picked up by compactor processes.  Its
# important that the instance props exactly match those of all other Accumulo
# processes in the cluster in order for them to be able to communicate with each
# other.  So if all other processes use ect2-0 as the zookeeper host, we could
# not use the IP addr here and still authenticate with other proccesses because
# the values for the prop would not match exactly.  See the comment about
# hostnames further down.
apiVersion: v1
kind: ConfigMap
metadata:
  name: accumulo-properties
data:
  accumulo.properties: |
    coordinator.port.client=30100
    general.rpc.timeout=240s
    instance.secret=muchos
    instance.volumes=hdfs://accucluster/accumulo
    instance.zookeeper.host=ect2-0:2181
---
apiVersion: v1
kind: ConfigMap
metadata:
  name: accumulo-logging
data:
  log4j2-service.properties: |
    status = info
    dest = err
    name = AccumuloCompactorLoggingProperties
    appender.console.type = Console
    appender.console.name = STDOUT
    appender.console.target = SYSTEM_OUT
    appender.console.layout.type = PatternLayout
    appender.console.layout.pattern = %d{ISO8601} [%-8c{2}] %-5p: %m%n
    appender.console.filter.threshold.type = ThresholdFilter
    appender.console.filter.threshold.level = info
    logger.hadoop.name = org.apache.hadoop
    logger.hadoop.level = debug
    logger.zookeeper.name = org.apache.zookeeper
    logger.zookeeper.level = error
    logger.accumulo.name = org.apache.accumulo
    logger.accumulo.level = debug
    rootLogger.level = info
    rootLogger.appenderRef.console.ref = STDOUT
---
# Muchos configured a nameservice of accucluster for HDFS.  Urls sent to
# compactors from the coordinator would be of the form hdfs://accucluster/... .
# Therefore compactors need to be able to resolve the accucluster nameservice.
# The following data is mapped into the containers as the hdfs-site.xml file in
# order to support resolving accucluster.
apiVersion: v1
kind: ConfigMap
metadata:
  name: hdfs-site
data:
  hdfs-site.xml: |
    <configuration>
      <property>
        <name>dfs.datanode.synconclose</name>
        <value>true</value>
      </property>
      <property>
        <name>dfs.nameservices</name>
        <value>accucluster</value>
      </property>
      <property>
        <name>dfs.internal.nameservices</name>
        <value>accucluster</value>
      </property>
      <property>
        <name>dfs.namenode.secondary.http-address.accucluster</name>
        <value>ect2-0:50090</value>
      </property>
      <property>
        <name>dfs.namenode.secondary.https-address.accucluster</name>
        <value>ect2-0:50091</value>
      </property>
      <property>
  	<name>dfs.namenode.rpc-address.accucluster</name>
        <value>ect2-0:8020</value>
      </property>
      <property>
        <name>dfs.namenode.http-address.accucluster</name>
        <value>ect2-0:50070</value>
      </property>
      <property>
        <name>dfs.namenode.https-address.accucluster</name>
        <value>ect2-0:50071</value>
      </property>
      <property>
        <name>dfs.namenode.servicerpc-address.accucluster</name>
        <value>ect2-0:8025</value>
      </property>
      <property>
        <name>dfs.namenode.lifeline.rpc-address.accucluster</name>
        <value>ect2-0:8050</value>
      </property>
      <property>
        <name>dfs.client.failover.proxy.provider.accucluster</name>
        <value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
       </property>
    </configuration>
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: accumulo-compactor
  labels:
    app: accumulo-compactor
spec:
  replicas: 10
  selector:
    matchLabels:
      app: accumulo-compactor
  template:
    metadata:
      labels:
        app: accumulo-compactor
    spec:
      # There was no shared DNS setup between the Muchos cluster and Kubernetes
      # cluster, however they did share a private vnet.  Also Kubernetes was
      # configured to give each pod an IP addr on the vnet, making the pods IP
      # addrs visible to any process on the Muchos cluster.  So the two
      # clusters could see each others IP addrs, but did not share a common
      # DNS.  Compactors need to access Zookeeper and the Coordinator and both
      # of these ran on host ect2-0, a hostname defined on the Muchos cluster.
      # Therefore the hostname ect2-0 is defined for each pod below so that it
      # can resolve this VM running in the Muchos cluster.  Given time a
      # solution using DNS between the two clusters could probably be figured
      # out.  This solution would probably be specific to an environment and
      # Muchos tries to make very few assumptions.
      hostNetwork: false
      hostAliases:
      - ip: "10.1.0.5"
        hostnames:
        - "ect2-0"
      containers:
      - name: accumulo-compactor
        image: extcompne.azurecr.io/accumulo
        command: ["/bin/bash","-c"]
        # Since the Kubernetes and Muchos clusters do not share a common DNS,
        # its very important that compactor processes advertise themselves in
        # Zookeeper using IP addrs.  In the following command -a $(hostname -I)
        # accomplishes this.
        args: ["accumulo compactor -q DCQ1 -a $(hostname -I)"]
        ports:
        - containerPort: 9101
        resources:
          requests:
            cpu: .9
            memory: "2048Mi"
          limits:
            cpu: 1
            memory: "2048Mi"
        env:
        - name: HADOOP_USER_NAME
          value: centos
        - name: ACCUMULO_JAVA_OPTS
          value: "-Xmx1g"
        volumeMounts:
        - name: "config"
          mountPath: "/opt/accumulo/conf/accumulo.properties"
          subPath: "accumulo.properties"
        - name: "logging"
          mountPath: "/opt/accumulo/conf/log4j2-service.properties"
          subPath: "log4j2-service.properties"
        - name: "hdfs"
          mountPath: "/opt/hadoop/etc/hadoop/hdfs-site.xml"
          subPath: "hdfs-site.xml"
      volumes:
      - name: "config"
        configMap:
          name: "accumulo-properties"
      - name: "logging"
        configMap:
          name: "accumulo-logging"
      - name: "hdfs"
        configMap:
          name: "hdfs-site"