contracts,notes,mapper

clusters/yc-k8s-test/infrastructure/bootstrap-jobs/service-bootstrap-jobs.yaml

@@ -10,6 +10,8 @@ spec:
   concurrencyPolicy: Forbid
   jobTemplate:
     spec:
+      backoffLimit: 0
+      activeDeadlineSeconds: 600
       template:
         metadata:
           annotations:
@@ -246,20 +248,20 @@ spec:
                   inter_broker_password="$(echo "${bootstrap_json}" | jq -r '.data.data.interBrokerPassword')"
                   [ -n "${inter_broker_password}" ] && [ "${inter_broker_password}" != "null" ]
                   list_json="$(curl -sS -H "X-Vault-Token: ${VAULT_TOKEN}" "${VAULT_ADDR}/v1/secrets/metadata/kafka/apps?list=true")"
-                  target_pod="$(kubectl -n kafka get pod kafka-kafka-contour-controller-0 -o jsonpath='{.metadata.name}' 2>/dev/null || true)"
+                  target_pod="kafka-kafka-contour-controller-0"
-                  if [ -z "${target_pod}" ]; then
+                  if ! kubectl -n kafka get pod "${target_pod}" >/dev/null 2>&1; then
-                    target_pod="$(kubectl -n kafka get pod -l app.kubernetes.io/component=controller-eligible -o jsonpath='{.items[0].metadata.name}' 2>/dev/null || true)"
-                  fi
-                  if [ -z "${target_pod}" ]; then
                     echo "Kafka controller pod not found"
                     exit 1
                   fi
+                  target_bootstrap="${target_pod}.kafka-kafka-contour-controller-headless.kafka.svc.cluster.local:9094"
 
                   admin_props="$(mktemp)"
                   printf "%s\n" \
                     "security.protocol=SASL_PLAINTEXT" \
                     "sasl.mechanism=PLAIN" \
                     "sasl.jaas.config=org.apache.kafka.common.security.plain.PlainLoginModule required username='inter_broker_user' password='${inter_broker_password}';" \
+                    "default.api.timeout.ms=60000" \
+                    "request.timeout.ms=60000" \
                     > "${admin_props}"
                   kubectl -n kafka exec "${target_pod}" -c kafka -- /bin/bash -lc 'cat > /tmp/admin.properties' < "${admin_props}"
                   rm -f "${admin_props}"
@@ -272,11 +274,25 @@ spec:
                     [ -z "${password}" ] && continue
 
                     echo "Reconciling Kafka user ${username}"
-                    kubectl -n kafka exec "${target_pod}" -c kafka -- /bin/bash -lc "\
+                    user_reconciled=false
-                      /opt/bitnami/kafka/bin/kafka-configs.sh --bootstrap-server localhost:9094 --command-config /tmp/admin.properties \
+                    attempt=1
-                        --alter --add-config 'SCRAM-SHA-512=[password=${password}]' \
+                    while [ "${attempt}" -le 3 ]; do
-                        --entity-type users --entity-name '${username}'
+                      if kubectl -n kafka exec "${target_pod}" -c kafka -- /bin/bash -lc "\
-                    " >/dev/null
+                        timeout 60 /opt/bitnami/kafka/bin/kafka-configs.sh --bootstrap-server ${target_bootstrap} --command-config /tmp/admin.properties \
+                          --alter --add-config 'SCRAM-SHA-512=[password=${password}]' \
+                          --entity-type users --entity-name '${username}'
+                      " >/dev/null; then
+                        user_reconciled=true
+                        break
+                      fi
+                      echo "Kafka user ${username} reconcile attempt ${attempt}/3 failed"
+                      attempt=$((attempt + 1))
+                      sleep 5
+                    done
+                    if [ "${user_reconciled}" != "true" ]; then
+                      echo "Kafka user ${username} reconcile failed, continue"
+                      continue
+                    fi
 
                     echo "${app_json}" | jq -c '.data.data.topics[]?' | while read -r topic_item; do
                       topic_name="$(echo "${topic_item}" | jq -r '.name // empty')"
@@ -286,16 +302,21 @@ spec:
                       [ -z "${topic_name}" ] && continue
 
                       echo "Reconciling Kafka topic ${topic_name}"
-                      kubectl -n kafka exec "${target_pod}" -c kafka -- /bin/bash -lc "\
+                      if ! kubectl -n kafka exec "${target_pod}" -c kafka -- /bin/bash -lc "\
-                        /opt/bitnami/kafka/bin/kafka-topics.sh --bootstrap-server localhost:9094 --command-config /tmp/admin.properties \
+                        timeout 40 /opt/bitnami/kafka/bin/kafka-topics.sh --bootstrap-server ${target_bootstrap} --command-config /tmp/admin.properties \
                           --create --if-not-exists --topic '${topic_name}' --partitions '${partitions}' --replication-factor '${replication_factor}'
-                      " >/dev/null
+                      " >/dev/null; then
+                        echo "Kafka topic ${topic_name} create/reconcile failed, continue"
+                        continue
+                      fi
 
                       if [ -n "${topic_configs}" ]; then
-                        kubectl -n kafka exec "${target_pod}" -c kafka -- /bin/bash -lc "\
+                        if ! kubectl -n kafka exec "${target_pod}" -c kafka -- /bin/bash -lc "\
-                          /opt/bitnami/kafka/bin/kafka-configs.sh --bootstrap-server localhost:9094 --command-config /tmp/admin.properties \
+                          timeout 40 /opt/bitnami/kafka/bin/kafka-configs.sh --bootstrap-server ${target_bootstrap} --command-config /tmp/admin.properties \
                             --alter --entity-type topics --entity-name '${topic_name}' --add-config '${topic_configs}'
-                        " >/dev/null
+                        " >/dev/null; then
+                          echo "Kafka topic ${topic_name} config reconcile failed, continue"
+                        fi
                       fi
                     done
                   done