Merge "Use TLS Zookeeper connections"
diff --git a/k8s/nodepool.yaml b/k8s/nodepool.yaml
index f3b2eab..eb5a7fe 100644
--- a/k8s/nodepool.yaml
+++ b/k8s/nodepool.yaml
@@ -34,8 +34,14 @@
volumeMounts:
- name: nodepool-config
mountPath: /etc/nodepool
+ - name: zookeeper-client-tls
+ mountPath: /tls/client
+ readOnly: true
volumes:
- name: nodepool-config
secret:
secretName: nodepool-gcs
+ - name: zookeeper-client-tls
+ secret:
+ secretName: zookeeper-client-tls
serviceAccountName: nodepool
diff --git a/k8s/zookeeper/README b/k8s/zookeeper/README
new file mode 100644
index 0000000..c0226d4
--- /dev/null
+++ b/k8s/zookeeper/README
@@ -0,0 +1,11 @@
+This directory contains deployment code for Zookeeper.
+
+The file `certs.yaml` is manually created and contains instructions
+for creating TLS certs for Zookeeper.
+
+The file `zookeeper.yaml` is the output of the Zookeeper Helm chart
+with https://github.com/helm/charts/pull/23480 applied and
+`values.yaml` as input.
+
+Once that PR is merged, we can switch to using the plain upstream Helm
+chart.
diff --git a/k8s/zookeeper/certs.yaml b/k8s/zookeeper/certs.yaml
new file mode 100644
index 0000000..b47fbf1
--- /dev/null
+++ b/k8s/zookeeper/certs.yaml
@@ -0,0 +1,25 @@
+---
+apiVersion: v1
+kind: Namespace
+metadata:
+ name: zookeeper
+---
+apiVersion: cert-manager.io/v1alpha2
+kind: Certificate
+metadata:
+ name: zookeeper-server
+ namespace: zookeeper
+spec:
+ keyEncoding: pkcs8
+ secretName: zookeeper-server-tls
+ commonName: server
+ dnsNames:
+ - zookeeper-0.zookeeper-headless.zookeeper.svc.cluster.local
+ - zookeeper-0
+ - zookeeper-1.zookeeper-headless.zookeeper.svc.cluster.local
+ - zookeeper-1
+ - zookeeper-2.zookeeper-headless.zookeeper.svc.cluster.local
+ - zookeeper-2
+ issuerRef:
+ name: ca-issuer
+ kind: ClusterIssuer
diff --git a/k8s/zookeeper/values.yaml b/k8s/zookeeper/values.yaml
new file mode 100644
index 0000000..336782b
--- /dev/null
+++ b/k8s/zookeeper/values.yaml
@@ -0,0 +1,313 @@
+## As weighted quorums are not supported, it is imperative that an odd number of replicas
+## be chosen. Moreover, the number of replicas should be either 1, 3, 5, or 7.
+##
+## ref: https://github.com/kubernetes/contrib/tree/master/statefulsets/zookeeper#stateful-set
+replicaCount: 3 # Desired quantity of ZooKeeper pods. This should always be (1,3,5, or 7)
+
+podDisruptionBudget:
+ maxUnavailable: 1 # Limits how many Zokeeper pods may be unavailable due to voluntary disruptions.
+
+terminationGracePeriodSeconds: 1800 # Duration in seconds a Zokeeper pod needs to terminate gracefully.
+
+updateStrategy:
+ type: RollingUpdate
+
+## refs:
+## - https://github.com/kubernetes/contrib/tree/master/statefulsets/zookeeper
+## - https://github.com/kubernetes/contrib/blob/master/statefulsets/zookeeper/Makefile#L1
+image:
+ repository: zookeeper # Container image repository for zookeeper container.
+ tag: 3.5.5 # Container image tag for zookeeper container.
+ pullPolicy: IfNotPresent # Image pull criteria for zookeeper container.
+
+service:
+ type: ClusterIP # Exposes zookeeper on a cluster-internal IP.
+ annotations: {} # Arbitrary non-identifying metadata for zookeeper service.
+ ## AWS example for use with LoadBalancer service type.
+ # external-dns.alpha.kubernetes.io/hostname: zookeeper.cluster.local
+ # service.beta.kubernetes.io/aws-load-balancer-cross-zone-load-balancing-enabled: "true"
+ # service.beta.kubernetes.io/aws-load-balancer-internal: "true"
+ ports:
+ client:
+ port: 2281 # Service port number for client port.
+ targetPort: client # Service target port for client port.
+ protocol: TCP # Service port protocol for client port.
+
+## Headless service.
+##
+headless:
+ annotations: {}
+ # publishNotReadyAddresses, default false for backward compatibility
+ # set to true to register DNS entries for unready pods, which helps in rare
+ # occasions when cluster is unable to be created, DNS caching is enforced
+ # or pods are in persistent crash loop
+ publishNotReadyAddresses: true
+
+ports:
+ client:
+ containerPort: 2281 # Port number for zookeeper container client port.
+ protocol: TCP # Protocol for zookeeper container client port.
+ election:
+ containerPort: 3888 # Port number for zookeeper container election port.
+ protocol: TCP # Protocol for zookeeper container election port.
+ server:
+ containerPort: 2888 # Port number for zookeeper container server port.
+ protocol: TCP # Protocol for zookeeper container server port.
+
+resources: {} # Optionally specify how much CPU and memory (RAM) each zookeeper container needs.
+ # We usually recommend not to specify default resources and to leave this as a conscious
+ # choice for the user. This also increases chances charts run on environments with little
+ # resources, such as Minikube. If you do want to specify resources, uncomment the following
+ # lines, adjust them as necessary, and remove the curly braces after 'resources:'.
+ # limits:
+ # cpu: 100m
+ # memory: 128Mi
+ # requests:
+ # cpu: 100m
+ # memory: 128Mi
+
+priorityClassName: ""
+
+nodeSelector: {} # Node label-values required to run zookeeper pods.
+
+tolerations: [] # Node taint overrides for zookeeper pods.
+
+affinity: {} # Criteria by which pod label-values influence scheduling for zookeeper pods.
+ # podAntiAffinity:
+ # requiredDuringSchedulingIgnoredDuringExecution:
+ # - topologyKey: "kubernetes.io/hostname"
+ # labelSelector:
+ # matchLabels:
+ # release: zookeeper
+
+podAnnotations: {} # Arbitrary non-identifying metadata for zookeeper pods.
+ # prometheus.io/scrape: "true"
+ # prometheus.io/path: "/metrics"
+ # prometheus.io/port: "9141"
+
+podLabels: {} # Key/value pairs that are attached to zookeeper pods.
+ # team: "developers"
+ # service: "zookeeper"
+
+securityContext:
+ fsGroup: 1000
+ runAsUser: 1000
+
+## Useful, if you want to use an alternate image.
+command:
+ - /bin/bash
+ - -xec
+ - /config-scripts/run
+
+## If you want the quorum to use SSL internally, create a
+## kubernetes.io/tls object (cert-manager produces these), and ensure
+## that it has a SAN entry for each of the server hostnames (e.g.:
+## zookeeper-0.zookeeper-headless.namespace.svc.cluster.local).
+## Supply the name of the object here:
+serverTlsSecret: "zookeeper-server-tls"
+
+## To use client SSL, create a kubernetes.io/tls object (cert-manager
+## produces these), and supply the name here. If this is provided,
+## then plain-text client connections are disabled. The same secret
+## may be used for both client and server connections.
+clientTlsSecret: "zookeeper-server-tls"
+
+## Useful if using any custom authorizer.
+## Pass any secrets to the kafka pods. Each secret will be passed as an
+## environment variable by default. The secret can also be mounted to a
+## specific path (in addition to environment variable) if required. Environment
+## variable names are generated as: `<secretName>_<secretKey>` (All upper case)
+# secrets:
+# - name: myKafkaSecret
+# keys:
+# - username
+# - password
+# # mountPath: /opt/kafka/secret
+# - name: myZkSecret
+# keys:
+# - user
+# - pass
+# mountPath: /opt/zookeeper/secret
+
+persistence:
+ enabled: true
+ ## zookeeper data Persistent Volume Storage Class
+ ## If defined, storageClassName: <storageClass>
+ ## If set to "-", storageClassName: "", which disables dynamic provisioning
+ ## If undefined (the default) or set to null, no storageClassName spec is
+ ## set, choosing the default provisioner. (gp2 on AWS, standard on
+ ## GKE, AWS & OpenStack)
+ ##
+ # storageClass: "-"
+ accessMode: ReadWriteOnce
+ size: 5Gi
+
+## Exporters query apps for metrics and make those metrics available for
+## Prometheus to scrape.
+exporters:
+
+ jmx:
+ enabled: false
+ image:
+ repository: sscaling/jmx-prometheus-exporter
+ tag: 0.3.0
+ pullPolicy: IfNotPresent
+ config:
+ lowercaseOutputName: false
+ ## ref: https://github.com/prometheus/jmx_exporter/blob/master/example_configs/zookeeper.yaml
+ rules:
+ - pattern: "org.apache.ZooKeeperService<name0=ReplicatedServer_id(\\d+)><>(\\w+)"
+ name: "zookeeper_$2"
+ - pattern: "org.apache.ZooKeeperService<name0=ReplicatedServer_id(\\d+), name1=replica.(\\d+)><>(\\w+)"
+ name: "zookeeper_$3"
+ labels:
+ replicaId: "$2"
+ - pattern: "org.apache.ZooKeeperService<name0=ReplicatedServer_id(\\d+), name1=replica.(\\d+), name2=(\\w+)><>(\\w+)"
+ name: "zookeeper_$4"
+ labels:
+ replicaId: "$2"
+ memberType: "$3"
+ - pattern: "org.apache.ZooKeeperService<name0=ReplicatedServer_id(\\d+), name1=replica.(\\d+), name2=(\\w+), name3=(\\w+)><>(\\w+)"
+ name: "zookeeper_$4_$5"
+ labels:
+ replicaId: "$2"
+ memberType: "$3"
+ startDelaySeconds: 30
+ env: {}
+ resources: {}
+ path: /metrics
+ ports:
+ jmxxp:
+ containerPort: 9404
+ protocol: TCP
+ livenessProbe:
+ httpGet:
+ path: /metrics
+ port: jmxxp
+ initialDelaySeconds: 30
+ periodSeconds: 15
+ timeoutSeconds: 60
+ failureThreshold: 8
+ successThreshold: 1
+ readinessProbe:
+ httpGet:
+ path: /metrics
+ port: jmxxp
+ initialDelaySeconds: 30
+ periodSeconds: 15
+ timeoutSeconds: 60
+ failureThreshold: 8
+ successThreshold: 1
+ serviceMonitor:
+ interval: 30s
+ scrapeTimeout: 30s
+ scheme: http
+
+ zookeeper:
+ ## refs:
+ ## - https://github.com/carlpett/zookeeper_exporter
+ ## - https://hub.docker.com/r/josdotso/zookeeper-exporter/
+ ## - https://www.datadoghq.com/blog/monitoring-kafka-performance-metrics/#zookeeper-metrics
+ enabled: false
+ image:
+ repository: josdotso/zookeeper-exporter
+ tag: v1.1.2
+ pullPolicy: IfNotPresent
+ config:
+ logLevel: info
+ resetOnScrape: "true"
+ env: {}
+ resources: {}
+ path: /metrics
+ ports:
+ zookeeperxp:
+ containerPort: 9141
+ protocol: TCP
+ livenessProbe:
+ httpGet:
+ path: /metrics
+ port: zookeeperxp
+ initialDelaySeconds: 30
+ periodSeconds: 15
+ timeoutSeconds: 60
+ failureThreshold: 8
+ successThreshold: 1
+ readinessProbe:
+ httpGet:
+ path: /metrics
+ port: zookeeperxp
+ initialDelaySeconds: 30
+ periodSeconds: 15
+ timeoutSeconds: 60
+ failureThreshold: 8
+ successThreshold: 1
+ serviceMonitor:
+ interval: 30s
+ scrapeTimeout: 30s
+ scheme: http
+
+## ServiceMonitor configuration in case you are using Prometheus Operator
+prometheus:
+ serviceMonitor:
+ ## If true a ServiceMonitor for each enabled exporter will be installed
+ enabled: false
+ ## The namespace where the ServiceMonitor(s) will be installed
+ # namespace: monitoring
+ ## The selector the Prometheus instance is searching for
+ ## [Default Prometheus Operator selector] (https://github.com/helm/charts/blob/f5a751f174263971fafd21eee4e35416d6612a3d/stable/prometheus-operator/templates/prometheus/prometheus.yaml#L74)
+ selector: {}
+
+## Use an alternate scheduler, e.g. "stork".
+## ref: https://kubernetes.io/docs/tasks/administer-cluster/configure-multiple-schedulers/
+##
+# schedulerName:
+
+## ref: https://github.com/kubernetes/contrib/tree/master/statefulsets/zookeeper
+env:
+
+ ## Options related to JMX exporter.
+ ## ref: https://github.com/apache/zookeeper/blob/master/bin/zkServer.sh#L36
+ JMXAUTH: "false"
+ JMXDISABLE: "false"
+ JMXPORT: 1099
+ JMXSSL: "false"
+
+ ## The port on which the server will accept client requests.
+ ZOO_PORT: 2181
+
+ ## The number of Ticks that an ensemble member is allowed to perform leader
+ ## election.
+ ZOO_INIT_LIMIT: 5
+
+ ZOO_TICK_TIME: 2000
+
+ ## The maximum number of concurrent client connections that
+ ## a server in the ensemble will accept.
+ ZOO_MAX_CLIENT_CNXNS: 60
+
+ ## The number of Tick by which a follower may lag behind the ensembles leader.
+ ZK_SYNC_LIMIT: 10
+
+ ## The number of wall clock ms that corresponds to a Tick for the ensembles
+ ## internal time.
+ ZK_TICK_TIME: 2000
+
+ ZOO_AUTOPURGE_PURGEINTERVAL: 0
+ ZOO_AUTOPURGE_SNAPRETAINCOUNT: 3
+ ZOO_STANDALONE_ENABLED: false
+
+jobs:
+ ## ref: http://zookeeper.apache.org/doc/r3.4.10/zookeeperProgrammers.html#ch_zkSessions
+ chroots:
+ enabled: false
+ activeDeadlineSeconds: 300
+ backoffLimit: 5
+ completions: 1
+ config:
+ create: []
+ # - /kafka
+ # - /ureplicator
+ env: []
+ parallelism: 1
+ resources: {}
+ restartPolicy: Never
diff --git a/k8s/zookeeper/zookeeper.yaml b/k8s/zookeeper/zookeeper.yaml
new file mode 100644
index 0000000..b61fce6
--- /dev/null
+++ b/k8s/zookeeper/zookeeper.yaml
@@ -0,0 +1,349 @@
+---
+# Source: zookeeper/templates/poddisruptionbudget.yaml
+apiVersion: policy/v1beta1
+kind: PodDisruptionBudget
+metadata:
+ name: zookeeper
+ labels:
+ app: zookeeper
+ chart: zookeeper-2.1.5
+ release: zookeeper
+ heritage: Helm
+ component: server
+spec:
+ selector:
+ matchLabels:
+ app: zookeeper
+ release: zookeeper
+ component: server
+ maxUnavailable: 1
+---
+# Source: zookeeper/templates/config-script.yaml
+apiVersion: v1
+kind: ConfigMap
+metadata:
+ name: zookeeper
+ labels:
+ app: zookeeper
+ chart: zookeeper-2.1.5
+ release: zookeeper
+ heritage: Helm
+ component: server
+data:
+ ok: |
+ #!/bin/sh
+ if [ -f /tls/client/ca.crt ]; then
+ echo "srvr" | openssl s_client -CAfile /tls/client/ca.crt -cert /tls/client/tls.crt -key /tls/client/tls.key -connect 127.0.0.1:${1:-2281} -quiet -ign_eof 2>/dev/null | grep Mode
+ else
+ zkServer.sh status
+ fi
+
+ ready: |
+ #!/bin/sh
+ if [ -f /tls/client/ca.crt ]; then
+ echo "ruok" | openssl s_client -CAfile /tls/client/ca.crt -cert /tls/client/tls.crt -key /tls/client/tls.key -connect 127.0.0.1:${1:-2281} -quiet -ign_eof 2>/dev/null
+ else
+ echo ruok | nc 127.0.0.1 ${1:-2181}
+ fi
+
+ run: |
+ #!/bin/bash
+
+ set -a
+ ROOT=$(echo /apache-zookeeper-*)
+
+ ZK_USER=${ZK_USER:-"zookeeper"}
+ ZK_LOG_LEVEL=${ZK_LOG_LEVEL:-"INFO"}
+ ZK_DATA_DIR=${ZK_DATA_DIR:-"/data"}
+ ZK_DATA_LOG_DIR=${ZK_DATA_LOG_DIR:-"/data/log"}
+ ZK_CONF_DIR=${ZK_CONF_DIR:-"/conf"}
+ ZK_CLIENT_PORT=${ZK_CLIENT_PORT:-2181}
+ ZK_SSL_CLIENT_PORT=${ZK_SSL_CLIENT_PORT:-2281}
+ ZK_SERVER_PORT=${ZK_SERVER_PORT:-2888}
+ ZK_ELECTION_PORT=${ZK_ELECTION_PORT:-3888}
+ ZK_TICK_TIME=${ZK_TICK_TIME:-2000}
+ ZK_INIT_LIMIT=${ZK_INIT_LIMIT:-10}
+ ZK_SYNC_LIMIT=${ZK_SYNC_LIMIT:-5}
+ ZK_HEAP_SIZE=${ZK_HEAP_SIZE:-2G}
+ ZK_MAX_CLIENT_CNXNS=${ZK_MAX_CLIENT_CNXNS:-60}
+ ZK_MIN_SESSION_TIMEOUT=${ZK_MIN_SESSION_TIMEOUT:- $((ZK_TICK_TIME*2))}
+ ZK_MAX_SESSION_TIMEOUT=${ZK_MAX_SESSION_TIMEOUT:- $((ZK_TICK_TIME*20))}
+ ZK_SNAP_RETAIN_COUNT=${ZK_SNAP_RETAIN_COUNT:-3}
+ ZK_PURGE_INTERVAL=${ZK_PURGE_INTERVAL:-0}
+ ID_FILE="$ZK_DATA_DIR/myid"
+ ZK_CONFIG_FILE="$ZK_CONF_DIR/zoo.cfg"
+ LOG4J_PROPERTIES="$ZK_CONF_DIR/log4j.properties"
+ HOST=$(hostname)
+ DOMAIN=`hostname -d`
+ JVMFLAGS="-Xmx$ZK_HEAP_SIZE -Xms$ZK_HEAP_SIZE"
+
+ APPJAR=$(echo $ROOT/*jar)
+ CLASSPATH="${ROOT}/lib/*:${APPJAR}:${ZK_CONF_DIR}:"
+
+ if [[ $HOST =~ (.*)-([0-9]+)$ ]]; then
+ NAME=${BASH_REMATCH[1]}
+ ORD=${BASH_REMATCH[2]}
+ MY_ID=$((ORD+1))
+ else
+ echo "Failed to extract ordinal from hostname $HOST"
+ exit 1
+ fi
+
+ mkdir -p $ZK_DATA_DIR
+ mkdir -p $ZK_DATA_LOG_DIR
+ echo $MY_ID >> $ID_FILE
+
+ if [[ -f /tls/server/ca.crt ]]; then
+ cp /tls/server/ca.crt /data/server-ca.pem
+ cat /tls/server/tls.crt /tls/server/tls.key > /data/server.pem
+ fi
+ if [[ -f /tls/client/ca.crt ]]; then
+ cp /tls/client/ca.crt /data/client-ca.pem
+ cat /tls/client/tls.crt /tls/client/tls.key > /data/client.pem
+ fi
+
+ echo "dataDir=$ZK_DATA_DIR" >> $ZK_CONFIG_FILE
+ echo "dataLogDir=$ZK_DATA_LOG_DIR" >> $ZK_CONFIG_FILE
+ echo "tickTime=$ZK_TICK_TIME" >> $ZK_CONFIG_FILE
+ echo "initLimit=$ZK_INIT_LIMIT" >> $ZK_CONFIG_FILE
+ echo "syncLimit=$ZK_SYNC_LIMIT" >> $ZK_CONFIG_FILE
+ echo "maxClientCnxns=$ZK_MAX_CLIENT_CNXNS" >> $ZK_CONFIG_FILE
+ echo "minSessionTimeout=$ZK_MIN_SESSION_TIMEOUT" >> $ZK_CONFIG_FILE
+ echo "maxSessionTimeout=$ZK_MAX_SESSION_TIMEOUT" >> $ZK_CONFIG_FILE
+ echo "autopurge.snapRetainCount=$ZK_SNAP_RETAIN_COUNT" >> $ZK_CONFIG_FILE
+ echo "autopurge.purgeInterval=$ZK_PURGE_INTERVAL" >> $ZK_CONFIG_FILE
+ echo "4lw.commands.whitelist=*" >> $ZK_CONFIG_FILE
+
+ # Client TLS configuration
+ if [[ -f /tls/client/ca.crt ]]; then
+ echo "secureClientPort=$ZK_SSL_CLIENT_PORT" >> $ZK_CONFIG_FILE
+ echo "ssl.keyStore.location=/data/client.pem" >> $ZK_CONFIG_FILE
+ echo "ssl.trustStore.location=/data/client-ca.pem" >> $ZK_CONFIG_FILE
+ else
+ echo "clientPort=$ZK_CLIENT_PORT" >> $ZK_CONFIG_FILE
+ fi
+
+ # Server TLS configuration
+ if [[ -f /tls/server/ca.crt ]]; then
+ echo "serverCnxnFactory=org.apache.zookeeper.server.NettyServerCnxnFactory" >> $ZK_CONFIG_FILE
+ echo "sslQuorum=true" >> $ZK_CONFIG_FILE
+ echo "ssl.quorum.keyStore.location=/data/server.pem" >> $ZK_CONFIG_FILE
+ echo "ssl.quorum.trustStore.location=/data/server-ca.pem" >> $ZK_CONFIG_FILE
+ fi
+
+ for (( i=1; i<=$ZK_REPLICAS; i++ ))
+ do
+ echo "server.$i=$NAME-$((i-1)).$DOMAIN:$ZK_SERVER_PORT:$ZK_ELECTION_PORT" >> $ZK_CONFIG_FILE
+ done
+
+ rm -f $LOG4J_PROPERTIES
+
+ echo "zookeeper.root.logger=$ZK_LOG_LEVEL, CONSOLE" >> $LOG4J_PROPERTIES
+ echo "zookeeper.console.threshold=$ZK_LOG_LEVEL" >> $LOG4J_PROPERTIES
+ echo "zookeeper.log.threshold=$ZK_LOG_LEVEL" >> $LOG4J_PROPERTIES
+ echo "zookeeper.log.dir=$ZK_DATA_LOG_DIR" >> $LOG4J_PROPERTIES
+ echo "zookeeper.log.file=zookeeper.log" >> $LOG4J_PROPERTIES
+ echo "zookeeper.log.maxfilesize=256MB" >> $LOG4J_PROPERTIES
+ echo "zookeeper.log.maxbackupindex=10" >> $LOG4J_PROPERTIES
+ echo "zookeeper.tracelog.dir=$ZK_DATA_LOG_DIR" >> $LOG4J_PROPERTIES
+ echo "zookeeper.tracelog.file=zookeeper_trace.log" >> $LOG4J_PROPERTIES
+ echo "log4j.rootLogger=\${zookeeper.root.logger}" >> $LOG4J_PROPERTIES
+ echo "log4j.appender.CONSOLE=org.apache.log4j.ConsoleAppender" >> $LOG4J_PROPERTIES
+ echo "log4j.appender.CONSOLE.Threshold=\${zookeeper.console.threshold}" >> $LOG4J_PROPERTIES
+ echo "log4j.appender.CONSOLE.layout=org.apache.log4j.PatternLayout" >> $LOG4J_PROPERTIES
+ echo "log4j.appender.CONSOLE.layout.ConversionPattern=%d{ISO8601} [myid:%X{myid}] - %-5p [%t:%C{1}@%L] - %m%n" >> $LOG4J_PROPERTIES
+
+ if [ -n "$JMXDISABLE" ]
+ then
+ MAIN=org.apache.zookeeper.server.quorum.QuorumPeerMain
+ else
+ MAIN="-Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.port=$JMXPORT -Dcom.sun.management.jmxremote.authenticate=$JMXAUTH -Dcom.sun.management.jmxremote.ssl=$JMXSSL -Dzookeeper.jmx.log4j.disable=$JMXLOG4J org.apache.zookeeper.server.quorum.QuorumPeerMain"
+ fi
+
+ set -x
+ exec java -cp "$CLASSPATH" $JVMFLAGS $MAIN $ZK_CONFIG_FILE
+---
+# Source: zookeeper/templates/service-headless.yaml
+apiVersion: v1
+kind: Service
+metadata:
+ name: zookeeper-headless
+ labels:
+ app: zookeeper
+ chart: zookeeper-2.1.5
+ release: zookeeper
+ heritage: Helm
+spec:
+ clusterIP: None
+ publishNotReadyAddresses: true
+ ports:
+ - name: client
+ port: 2281
+ targetPort: client
+ protocol: TCP
+ - name: election
+ port: 3888
+ targetPort: election
+ protocol: TCP
+ - name: server
+ port: 2888
+ targetPort: server
+ protocol: TCP
+ selector:
+ app: zookeeper
+ release: zookeeper
+---
+# Source: zookeeper/templates/service.yaml
+apiVersion: v1
+kind: Service
+metadata:
+ name: zookeeper
+ labels:
+ app: zookeeper
+ chart: zookeeper-2.1.5
+ release: zookeeper
+ heritage: Helm
+spec:
+ type: ClusterIP
+ ports:
+ - name: client
+ port: 2281
+ protocol: TCP
+ targetPort: client
+ selector:
+ app: zookeeper
+ release: zookeeper
+---
+# Source: zookeeper/templates/statefulset.yaml
+apiVersion: apps/v1
+kind: StatefulSet
+metadata:
+ name: zookeeper
+ labels:
+ app: zookeeper
+ chart: zookeeper-2.1.5
+ release: zookeeper
+ heritage: Helm
+ component: server
+spec:
+ serviceName: zookeeper-headless
+ replicas: 3
+ selector:
+ matchLabels:
+ app: zookeeper
+ release: zookeeper
+ component: server
+ updateStrategy:
+ type: RollingUpdate
+ template:
+ metadata:
+ labels:
+ app: zookeeper
+ release: zookeeper
+ component: server
+ spec:
+ terminationGracePeriodSeconds: 1800
+ securityContext:
+ fsGroup: 1000
+ runAsUser: 1000
+ containers:
+
+ - name: zookeeper
+ image: "zookeeper:3.5.5"
+ imagePullPolicy: IfNotPresent
+ command:
+ - "/bin/bash"
+ - "-xec"
+ - "/config-scripts/run"
+ ports:
+ - name: client
+ containerPort: 2281
+ protocol: TCP
+ - name: election
+ containerPort: 3888
+ protocol: TCP
+ - name: server
+ containerPort: 2888
+ protocol: TCP
+ livenessProbe:
+ exec:
+ command:
+ - sh
+ - /config-scripts/ok
+ initialDelaySeconds: 20
+ periodSeconds: 30
+ timeoutSeconds: 5
+ failureThreshold: 2
+ successThreshold: 1
+ readinessProbe:
+ exec:
+ command:
+ - sh
+ - /config-scripts/ready
+ initialDelaySeconds: 20
+ periodSeconds: 30
+ timeoutSeconds: 5
+ failureThreshold: 2
+ successThreshold: 1
+ env:
+ - name: ZK_REPLICAS
+ value: "3"
+ - name: JMXAUTH
+ value: "false"
+ - name: JMXDISABLE
+ value: "false"
+ - name: JMXPORT
+ value: "1099"
+ - name: JMXSSL
+ value: "false"
+ - name: ZK_SYNC_LIMIT
+ value: "10"
+ - name: ZK_TICK_TIME
+ value: "2000"
+ - name: ZOO_AUTOPURGE_PURGEINTERVAL
+ value: "0"
+ - name: ZOO_AUTOPURGE_SNAPRETAINCOUNT
+ value: "3"
+ - name: ZOO_INIT_LIMIT
+ value: "5"
+ - name: ZOO_MAX_CLIENT_CNXNS
+ value: "60"
+ - name: ZOO_PORT
+ value: "2181"
+ - name: ZOO_STANDALONE_ENABLED
+ value: "false"
+ - name: ZOO_TICK_TIME
+ value: "2000"
+ resources:
+ {}
+ volumeMounts:
+ - name: data
+ mountPath: /data
+ - name: zookeeper-server-tls
+ mountPath: /tls/server
+ readOnly: true
+ - name: zookeeper-client-tls
+ mountPath: /tls/client
+ readOnly: true
+ - name: config
+ mountPath: /config-scripts
+ volumes:
+ - name: config
+ configMap:
+ name: zookeeper
+ defaultMode: 0555
+ - name: zookeeper-server-tls
+ secret:
+ secretName: zookeeper-server-tls
+ - name: zookeeper-client-tls
+ secret:
+ secretName: zookeeper-server-tls
+ volumeClaimTemplates:
+ - metadata:
+ name: data
+ spec:
+ accessModes:
+ - "ReadWriteOnce"
+ resources:
+ requests:
+ storage: "5Gi"
diff --git a/k8s/zuul.yaml b/k8s/zuul.yaml
index 1cd88e1..275039e 100644
--- a/k8s/zuul.yaml
+++ b/k8s/zuul.yaml
@@ -1,4 +1,17 @@
---
+apiVersion: cert-manager.io/v1alpha2
+kind: Certificate
+metadata:
+ name: zookeeper-client
+ namespace: zuul
+spec:
+ keyEncoding: pkcs8
+ secretName: zookeeper-client-tls
+ commonName: client
+ issuerRef:
+ name: ca-issuer
+ kind: ClusterIssuer
+---
apiVersion: cloud.google.com/v1beta1
kind: BackendConfig
metadata:
@@ -114,10 +127,16 @@
volumeMounts:
- name: zuul-config
mountPath: /etc/zuul
+ - name: zookeeper-client-tls
+ mountPath: /tls/client
+ readOnly: true
volumes:
- name: zuul-config
secret:
secretName: zuul-config
+ - name: zookeeper-client-tls
+ secret:
+ secretName: zookeeper-client-tls
---
apiVersion: apps/v1
kind: Deployment
@@ -157,10 +176,16 @@
volumeMounts:
- name: zuul-config
mountPath: /etc/zuul
+ - name: zookeeper-client-tls
+ mountPath: /tls/client
+ readOnly: true
volumes:
- name: zuul-config
secret:
secretName: zuul-config
+ - name: zookeeper-client-tls
+ secret:
+ secretName: zookeeper-client-tls
---
apiVersion: apps/v1
kind: StatefulSet
@@ -210,6 +235,9 @@
mountPath: /var/lib/zuul
- name: nodepool-private-key
mountPath: /var/lib/zuul/ssh
+ - name: zookeeper-client-tls
+ mountPath: /tls/client
+ readOnly: true
securityContext:
privileged: true
volumes:
@@ -225,6 +253,9 @@
- name: nodepool-private-key
secret:
secretName: nodepool-private-key
+ - name: zookeeper-client-tls
+ secret:
+ secretName: zookeeper-client-tls
---
apiVersion: apps/v1
kind: StatefulSet
@@ -269,6 +300,9 @@
mountPath: /etc/zuul/tenant
- name: zuul-scheduler
mountPath: /var/lib/zuul
+ - name: zookeeper-client-tls
+ mountPath: /tls/client
+ readOnly: true
volumes:
- name: zuul-config
secret:
@@ -276,6 +310,9 @@
- name: zuul-tenant-config
secret:
secretName: zuul-tenant-config
+ - name: zookeeper-client-tls
+ secret:
+ secretName: zookeeper-client-tls
serviceAccountName: zuul
volumeClaimTemplates:
- metadata:
diff --git a/nodepool/nodepool.yaml b/nodepool/nodepool.yaml
index b2f3a6e..6b2bc47 100644
--- a/nodepool/nodepool.yaml
+++ b/nodepool/nodepool.yaml
@@ -1,7 +1,11 @@
diskimages: []
zookeeper-servers:
- host: zookeeper.zookeeper
- port: 2181
+ port: 2281
+zookeeper-tls:
+ ca: /tls/client/ca.crt
+ cert: /tls/client/tls.crt
+ key: /tls/client/tls.key
providers:
- name: gcloud-provider
driver: gce
diff --git a/playbooks/deploy.yaml b/playbooks/deploy.yaml
index 073d283..9495de8 100644
--- a/playbooks/deploy.yaml
+++ b/playbooks/deploy.yaml
@@ -19,6 +19,11 @@
state: present
src: "{{ root }}/k8s/certmanager.yaml"
+ - name: Update Zookeeper deployment
+ k8s:
+ state: present
+ src: "{{ root }}/k8s/zookeeper/zookeeper.yaml"
+
- name: Update Letsencrypt configuration
k8s:
state: present
diff --git a/zuul/zuul.conf b/zuul/zuul.conf
index 31b34fc..e770e93 100644
--- a/zuul/zuul.conf
+++ b/zuul/zuul.conf
@@ -3,7 +3,10 @@
port=4730
[zookeeper]
-hosts=zookeeper.zookeeper
+hosts=zookeeper.zookeeper:2281
+tls_ca=/tls/client/ca.crt
+tls_cert=/tls/client/tls.crt
+tls_key=/tls/client/tls.key
[gearman_server]
start=true
