refactor pooler tls support and set pooler pod security context

.gitignore

@@ -95,6 +95,7 @@ coverage.xml
 
 # e2e tests
 e2e/manifests
+e2e/tls
 
 # Translations
 *.mo

charts/postgres-operator/crds/operatorconfigurations.yaml

@@ -637,7 +637,7 @@ spec:
                     default: "pooler"
                   connection_pooler_image:
                     type: string
-                    default: "registry.opensource.zalan.do/acid/pgbouncer:master-26"
+                    default: "registry.opensource.zalan.do/acid/pgbouncer:master-27"
                   connection_pooler_max_db_connections:
                     type: integer
                     default: 60

charts/postgres-operator/values.yaml

@@ -416,7 +416,7 @@ configConnectionPooler:
   # db user for pooler to use
   connection_pooler_user: "pooler"
   # docker image
-  connection_pooler_image: "registry.opensource.zalan.do/acid/pgbouncer:master-26"
+  connection_pooler_image: "registry.opensource.zalan.do/acid/pgbouncer:master-27"
   # max db connections the pooler should hold
   connection_pooler_max_db_connections: 60
   # default pooling mode

docs/reference/cluster_manifest.md

@@ -543,7 +543,9 @@ for both master and replica pooler services (if `enableReplicaConnectionPooler`
 
 ## Custom TLS certificates
 
-Those parameters are grouped under the `tls` top-level key.
+Those parameters are grouped under the `tls` top-level key. Note, you have to
+define `spiloFSGroup` in the Postgres cluster manifest or `spilo_fsgroup` in
+the global configuration before adding the `tls` section'.
 
 * **secretName**
   By setting the `secretName` value, the cluster will switch to load the given

docs/user.md

@@ -1197,14 +1197,19 @@ don't know the value, use `103` which is the GID from the default Spilo image
 OpenShift allocates the users and groups dynamically (based on scc), and their
 range is different in every namespace. Due to this dynamic behaviour, it's not
 trivial to know at deploy time the uid/gid of the user in the cluster.
-Therefore, instead of using a global `spilo_fsgroup` setting, use the
-`spiloFSGroup` field per Postgres cluster.
+Therefore, instead of using a global `spilo_fsgroup` setting in operator
+configuration or use the `spiloFSGroup` field per Postgres cluster manifest.
+
+For testing purposes, you can generate a self-signed certificate with openssl:
+```sh
+openssl req -x509 -nodes -newkey rsa:2048 -keyout tls.key -out tls.crt -subj "/CN=acid.zalan.do"
+```
 
 Upload the cert as a kubernetes secret:
 ```sh
 kubectl create secret tls pg-tls \
-  --key pg-tls.key \
-  --cert pg-tls.crt
+  --key tls.key \
+  --cert tls.crt
 ```
 
 When doing client auth, CA can come optionally from the same secret:
@@ -1231,8 +1236,7 @@ spec:
 
 Optionally, the CA can be provided by a different secret:
 ```sh
-kubectl create secret generic pg-tls-ca \
-  --from-file=ca.crt=ca.crt
+kubectl create secret generic pg-tls-ca --from-file=ca.crt=ca.crt
 ```
 
 Then configure the postgres resource with the TLS secret:
@@ -1255,3 +1259,16 @@ Alternatively, it is also possible to use
 
 Certificate rotation is handled in the Spilo image which checks every 5
 minutes if the certificates have changed and reloads postgres accordingly.
+
+### TLS certificates for connection pooler
+
+By default, the pgBouncer image generates its own TLS certificate like Spilo.
+When the `tls` section is specfied in the manifest it will be used for the
+connection pooler pod(s) as well. The security context options are hard coded
+to `runAsUser: 100` and `runAsGroup: 101`. The `fsGroup` will be the same
+like for Spilo.
+
+As of now, the operator does not sync the pooler deployment automatically
+which means that changes in the pod template are not caught. You need to
+toggle `enableConnectionPooler` to set environment variables, volumes, secret
+mounts and securityContext required for TLS support in the pooler pod.

e2e/Makefile

@@ -29,10 +29,12 @@ default: tools
 
 clean:
 	rm -rf manifests
+	rm -rf tls
 
 copy: clean
 	mkdir manifests
 	cp -r ../manifests .
+	mkdir tls
 
 docker: scm-source.json
 	docker build -t "$(IMAGE):$(TAG)" .

e2e/run.sh

@@ -55,13 +55,18 @@ function set_kind_api_server_ip(){
   sed -i "s/server.*$/server: https:\/\/$kind_api_server/g" "${kubeconfig_path}"
 }
 
+function generate_certificate(){
+  openssl req -x509 -nodes -newkey rsa:2048 -keyout tls/tls.key -out tls/tls.crt -subj "/CN=acid.zalan.do"
+}
+
 function run_tests(){
   echo "Running tests... image: ${e2e_test_runner_image}"
   # tests modify files in ./manifests, so we mount a copy of this directory done by the e2e Makefile
 
   docker run --rm --network=host -e "TERM=xterm-256color" \
   --mount type=bind,source="$(readlink -f ${kubeconfig_path})",target=/root/.kube/config \
   --mount type=bind,source="$(readlink -f manifests)",target=/manifests \
+  --mount type=bind,source="$(readlink -f tls)",target=/tls \
   --mount type=bind,source="$(readlink -f tests)",target=/tests \
   --mount type=bind,source="$(readlink -f exec.sh)",target=/exec.sh \
   --mount type=bind,source="$(readlink -f scripts)",target=/scripts \
@@ -82,6 +87,7 @@ function main(){
   [[ ! -f ${kubeconfig_path} ]] && start_kind
   load_operator_image
   set_kind_api_server_ip
+  generate_certificate
 
   shift
   run_tests $@

e2e/tests/k8s_api.py

@@ -156,6 +156,26 @@ def get_services():
         while not get_services():
             time.sleep(self.RETRY_TIMEOUT_SEC)
 
+    def count_pods_with_volume_mount(self, mount_name, labels, namespace='default'):
+        pod_count = 0
+        pods = self.api.core_v1.list_namespaced_pod(namespace, label_selector=labels).items
+        for pod in pods:
+            for mount in pod.spec.containers[0].volume_mounts:
+                if mount.name == mount_name:
+                    pod_count += 1
+
+        return pod_count
+
+    def count_pods_with_env_variable(self, env_variable_key, labels, namespace='default'):
+        pod_count = 0
+        pods = self.api.core_v1.list_namespaced_pod(namespace, label_selector=labels).items
+        for pod in pods:
+            for env in pod.spec.containers[0].env:
+                if env.name == env_variable_key:
+                    pod_count += 1
+
+        return pod_count
+
     def count_pods_with_rolling_update_flag(self, labels, namespace='default'):
         pods = self.api.core_v1.list_namespaced_pod(namespace, label_selector=labels).items
         return len(list(filter(lambda x: "zalando-postgres-operator-rolling-update-required" in x.metadata.annotations, pods)))
@@ -241,6 +261,18 @@ def update_config(self, config_map_patch, step="Updating operator deployment"):
     def patch_pod(self, data, pod_name, namespace="default"):
         self.api.core_v1.patch_namespaced_pod(pod_name, namespace, data)
 
+    def create_tls_secret_with_kubectl(self, secret_name):
+        return subprocess.run(
+            ["kubectl", "create", "secret", "tls", secret_name, "--key=tls/tls.key", "--cert=tls/tls.crt"],
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE)
+
+    def create_tls_ca_secret_with_kubectl(self, secret_name):
+        return subprocess.run(
+            ["kubectl", "create", "secret", "generic", secret_name, "--from-file=ca.crt=tls/ca.crt"],
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE)
+
     def create_with_kubectl(self, path):
         return subprocess.run(
             ["kubectl", "apply", "-f", path],

e2e/tests/test_e2e.py

@@ -622,6 +622,49 @@ def test_cross_namespace_secrets(self):
         self.eventuallyEqual(lambda: k8s.count_secrets_with_label("cluster-name=acid-minimal-cluster,application=spilo", self.test_namespace),
                              1, "Secret not created for user in namespace")
 
+    @timeout_decorator.timeout(TEST_TIMEOUT_SEC)
+    def test_custom_ssl_certificate(self):
+        '''
+        Test if spilo uses a custom SSL certificate
+        '''
+
+        k8s = self.k8s
+        cluster_label = 'application=spilo,cluster-name=acid-minimal-cluster'
+        tls_secret = "pg-tls"
+
+        # get nodes of master and replica(s) (expected target of new master)
+        _, replica_nodes = k8s.get_pg_nodes(cluster_label)
+        self.assertNotEqual(replica_nodes, [])
+
+        try:
+            # create secret containing ssl certificate
+            result = self.k8s.create_tls_secret_with_kubectl(tls_secret)
+            print("stdout: {}, stderr: {}".format(result.stdout, result.stderr))
+
+            # enable load balancer services
+            pg_patch_tls = {
+                "spec": {
+                    "spiloFSGroup": 103,
+                    "tls": {
+                        "secretName": tls_secret
+                    }
+                }
+            }
+            k8s.api.custom_objects_api.patch_namespaced_custom_object(
+                "acid.zalan.do", "v1", "default", "postgresqls", "acid-minimal-cluster", pg_patch_tls)
+
+            # wait for switched over
+            k8s.wait_for_pod_failover(replica_nodes, 'spilo-role=master,' + cluster_label)
+            k8s.wait_for_pod_start('spilo-role=replica,' + cluster_label)
+
+            self.eventuallyEqual(lambda: k8s.count_pods_with_env_variable("SSL_CERTIFICATE_FILE", cluster_label), 2, "TLS env variable SSL_CERTIFICATE_FILE missing in Spilo pods")
+            self.eventuallyEqual(lambda: k8s.count_pods_with_env_variable("SSL_PRIVATE_KEY_FILE", cluster_label), 2, "TLS env variable SSL_PRIVATE_KEY_FILE missing in Spilo pods")
+            self.eventuallyEqual(lambda: k8s.count_pods_with_volume_mount(tls_secret, cluster_label), 2, "TLS volume mount missing in Spilo pods")
+
+        except timeout_decorator.TimeoutError:
+            print('Operator log: {}'.format(k8s.get_operator_log()))
+            raise
+
     @timeout_decorator.timeout(TEST_TIMEOUT_SEC)
     def test_enable_disable_connection_pooler(self):
         '''
@@ -653,6 +696,11 @@ def test_enable_disable_connection_pooler(self):
         self.eventuallyEqual(lambda: k8s.count_services_with_label(pooler_label), 2, "No pooler service found")
         self.eventuallyEqual(lambda: k8s.count_secrets_with_label(pooler_label), 1, "Pooler secret not created")
 
+        # TLS still enabled so check existing env variables and volume mounts
+        self.eventuallyEqual(lambda: k8s.count_pods_with_env_variable("CONNECTION_POOLER_CLIENT_TLS_CRT", pooler_label), 4, "TLS env variable CONNECTION_POOLER_CLIENT_TLS_CRT missing in pooler pods")
+        self.eventuallyEqual(lambda: k8s.count_pods_with_env_variable("CONNECTION_POOLER_CLIENT_TLS_KEY", pooler_label), 4, "TLS env variable CONNECTION_POOLER_CLIENT_TLS_KEY missing in pooler pods")
+        self.eventuallyEqual(lambda: k8s.count_pods_with_volume_mount("pg-tls", pooler_label), 4, "TLS volume mount missing in pooler pods")
+
         k8s.api.custom_objects_api.patch_namespaced_custom_object(
             'acid.zalan.do', 'v1', 'default',
             'postgresqls', 'acid-minimal-cluster',

manifests/configmap.yaml

@@ -17,7 +17,7 @@ data:
   # connection_pooler_default_cpu_request: "500m"
   # connection_pooler_default_memory_limit: 100Mi
   # connection_pooler_default_memory_request: 100Mi
-  connection_pooler_image: "registry.opensource.zalan.do/acid/pgbouncer:master-26"
+  connection_pooler_image: "registry.opensource.zalan.do/acid/pgbouncer:master-27"
   # connection_pooler_max_db_connections: 60
   # connection_pooler_mode: "transaction"
   # connection_pooler_number_of_instances: 2

manifests/minimal-fake-pooler-deployment.yaml

@@ -23,7 +23,7 @@ spec:
       serviceAccountName: postgres-operator
       containers:
       - name: postgres-operator
-        image: registry.opensource.zalan.do/acid/pgbouncer:master-26
+        image: registry.opensource.zalan.do/acid/pgbouncer:master-27
         imagePullPolicy: IfNotPresent
         resources:
           requests:

manifests/operatorconfiguration.crd.yaml

@@ -635,7 +635,7 @@ spec:
                     default: "pooler"
                   connection_pooler_image:
                     type: string
-                    default: "registry.opensource.zalan.do/acid/pgbouncer:master-26"
+                    default: "registry.opensource.zalan.do/acid/pgbouncer:master-27"
                   connection_pooler_max_db_connections:
                     type: integer
                     default: 60

manifests/postgresql-operator-default-configuration.yaml

@@ -203,7 +203,7 @@ configuration:
     connection_pooler_default_cpu_request: "500m"
     connection_pooler_default_memory_limit: 100Mi
     connection_pooler_default_memory_request: 100Mi
-    connection_pooler_image: "registry.opensource.zalan.do/acid/pgbouncer:master-26"
+    connection_pooler_image: "registry.opensource.zalan.do/acid/pgbouncer:master-27"
     # connection_pooler_max_db_connections: 60
     connection_pooler_mode: "transaction"
     connection_pooler_number_of_instances: 2