Merge branch 'main' into feat/region-mover

Author: razvan

CHANGELOG.md

@@ -13,6 +13,7 @@
 
 - Implement `envOverrides` for HbaseCluster ([#550]).
 - Omid test: use 1.1.2, update default port number and raise test timeout ([#556]).
+- An invalid `HBaseCluster` doesn't cause the operator to stop functioning (#[575]).
 
 ### Removed
 

rust/operator-binary/src/hbase_controller.rs

@@ -29,6 +29,7 @@ use stackable_operator::{
         },
         apimachinery::pkg::{apis::meta::v1::LabelSelector, util::intstr::IntOrString},
     },
+    kube::core::{error_boundary, DeserializeGuard},
     kube::{runtime::controller::Action, Resource},
     kvp::{Label, LabelError, Labels, ObjectLabels},
     logging::controller::ReconcilerError,
@@ -280,6 +281,11 @@ pub enum Error {
 
     #[snafu(display("authorization is only supported from HBase 2.6 onwards"))]
     AuthorizationNotSupported,
+
+    #[snafu(display("HBaseCluster object is invalid"))]
+    InvalidHBaseCluster {
+        source: error_boundary::InvalidObject,
+    },
 }
 
 type Result<T, E = Error> = std::result::Result<T, E>;
@@ -290,31 +296,39 @@ impl ReconcilerError for Error {
     }
 }
 
-pub async fn reconcile_hbase(hbase: Arc<HbaseCluster>, ctx: Arc<Ctx>) -> Result<Action> {
+pub async fn reconcile_hbase(
+    hbase: Arc<DeserializeGuard<HbaseCluster>>,
+    ctx: Arc<Ctx>,
+) -> Result<Action> {
     tracing::info!("Starting reconcile");
 
+    let hbase = hbase
+        .0
+        .as_ref()
+        .map_err(error_boundary::InvalidObject::clone)
+        .context(InvalidHBaseClusterSnafu)?;
+
     let client = &ctx.client;
 
-    validate_cr(&hbase)?;
+    validate_cr(hbase)?;
 
     let resolved_product_image = hbase
         .spec
         .image
         .resolve(DOCKER_IMAGE_BASE_NAME, crate::built_info::PKG_VERSION);
-    let zookeeper_connection_information = ZookeeperConnectionInformation::retrieve(&hbase, client)
+    let zookeeper_connection_information = ZookeeperConnectionInformation::retrieve(hbase, client)
         .await
         .context(RetrieveZookeeperConnectionInformationSnafu)?;
 
-    let vector_aggregator_address = resolve_vector_aggregator_address(&hbase, client)
+    let vector_aggregator_address = resolve_vector_aggregator_address(hbase, client)
         .await
         .context(ResolveVectorAggregatorAddressSnafu)?;
 
     let roles = hbase.build_role_properties().context(RolePropertiesSnafu)?;
 
     let validated_config = validate_all_roles_and_groups_config(
         &resolved_product_image.app_version_label,
-        &transform_all_roles_to_config(hbase.as_ref(), roles)
-            .context(GenerateProductConfigSnafu)?,
+        &transform_all_roles_to_config(hbase, roles).context(GenerateProductConfigSnafu)?,
         &ctx.product_config,
         false,
         false,
@@ -323,7 +337,7 @@ pub async fn reconcile_hbase(hbase: Arc<HbaseCluster>, ctx: Arc<Ctx>) -> Result<
 
     let hbase_opa_config = match &hbase.spec.cluster_config.authorization {
         Some(opa_config) => Some(
-            HbaseOpaConfig::from_opa_config(client, &hbase, opa_config)
+            HbaseOpaConfig::from_opa_config(client, hbase, opa_config)
                 .await
                 .context(InvalidOpaConfigSnafu)?,
         ),
@@ -340,15 +354,15 @@ pub async fn reconcile_hbase(hbase: Arc<HbaseCluster>, ctx: Arc<Ctx>) -> Result<
     .context(CreateClusterResourcesSnafu)?;
 
     let region_server_role_service =
-        build_region_server_role_service(&hbase, &resolved_product_image)?;
+        build_region_server_role_service(hbase, &resolved_product_image)?;
     cluster_resources
         .add(client, region_server_role_service)
         .await
         .context(ApplyRoleServiceSnafu)?;
 
     // discovery config map
     let discovery_cm = build_discovery_configmap(
-        &hbase,
+        hbase,
         &zookeeper_connection_information,
         &resolved_product_image,
     )
@@ -359,7 +373,7 @@ pub async fn reconcile_hbase(hbase: Arc<HbaseCluster>, ctx: Arc<Ctx>) -> Result<
         .context(ApplyDiscoveryConfigMapSnafu)?;
 
     let (rbac_sa, rbac_rolebinding) = build_rbac_resources(
-        hbase.as_ref(),
+        hbase,
         APP_NAME,
         cluster_resources
             .get_required_labels()
@@ -393,9 +407,9 @@ pub async fn reconcile_hbase(hbase: Arc<HbaseCluster>, ctx: Arc<Ctx>) -> Result<
                 .context(FailedToResolveConfigSnafu)?;
 
             let rg_service =
-                build_rolegroup_service(&hbase, &hbase_role, &rolegroup, &resolved_product_image)?;
+                build_rolegroup_service(hbase, &hbase_role, &rolegroup, &resolved_product_image)?;
             let rg_configmap = build_rolegroup_config_map(
-                &hbase,
+                hbase,
                 &rolegroup,
                 rolegroup_config,
                 &zookeeper_connection_information,
@@ -405,7 +419,7 @@ pub async fn reconcile_hbase(hbase: Arc<HbaseCluster>, ctx: Arc<Ctx>) -> Result<
                 vector_aggregator_address.as_deref(),
             )?;
             let rg_statefulset = build_rolegroup_statefulset(
-                &hbase,
+                hbase,
                 &hbase_role,
                 &rolegroup,
                 rolegroup_config,
@@ -439,7 +453,7 @@ pub async fn reconcile_hbase(hbase: Arc<HbaseCluster>, ctx: Arc<Ctx>) -> Result<
             pod_disruption_budget: pdb,
         }) = role_config
         {
-            add_pdbs(pdb, &hbase, &hbase_role, client, &mut cluster_resources)
+            add_pdbs(pdb, hbase, &hbase_role, client, &mut cluster_resources)
                 .await
                 .context(FailedToCreatePdbSnafu)?;
         }
@@ -449,18 +463,15 @@ pub async fn reconcile_hbase(hbase: Arc<HbaseCluster>, ctx: Arc<Ctx>) -> Result<
         ClusterOperationsConditionBuilder::new(&hbase.spec.cluster_operation);
 
     let status = HbaseClusterStatus {
-        conditions: compute_conditions(
-            hbase.as_ref(),
-            &[&ss_cond_builder, &cluster_operation_cond_builder],
-        ),
+        conditions: compute_conditions(hbase, &[&ss_cond_builder, &cluster_operation_cond_builder]),
     };
 
     cluster_resources
         .delete_orphaned_resources(client)
         .await
         .context(DeleteOrphanedResourcesSnafu)?;
     client
-        .apply_patch_status(OPERATOR_NAME, hbase.as_ref(), &status)
+        .apply_patch_status(OPERATOR_NAME, hbase, &status)
         .await
         .context(ApplyStatusSnafu)?;
 
@@ -998,8 +1009,16 @@ where
     })
 }
 
-pub fn error_policy(_obj: Arc<HbaseCluster>, _error: &Error, _ctx: Arc<Ctx>) -> Action {
-    Action::requeue(*Duration::from_secs(5))
+pub fn error_policy(
+    _obj: Arc<DeserializeGuard<HbaseCluster>>,
+    error: &Error,
+    _ctx: Arc<Ctx>,
+) -> Action {
+    match error {
+        // root object is invalid, will be requed when modified
+        Error::InvalidHBaseCluster { .. } => Action::await_change(),
+        _ => Action::requeue(*Duration::from_secs(5)),
+    }
 }
 
 pub fn build_recommended_labels<'a>(

rust/operator-binary/src/main.rs

@@ -14,6 +14,7 @@ use stackable_hbase_crd::{HbaseCluster, APP_NAME};
 use stackable_operator::{
     cli::{Command, ProductOperatorRun},
     k8s_openapi::api::{apps::v1::StatefulSet, core::v1::Service},
+    kube::core::DeserializeGuard,
     kube::runtime::{controller::Controller, watcher},
     logging::controller::report_controller_reconciled,
     CustomResourceExt,
@@ -66,7 +67,7 @@ async fn main() -> anyhow::Result<()> {
                 stackable_operator::client::create_client(Some(OPERATOR_NAME.to_string())).await?;
 
             Controller::new(
-                watch_namespace.get_api::<HbaseCluster>(&client),
+                watch_namespace.get_api::<DeserializeGuard<HbaseCluster>>(&client),
                 watcher::Config::default(),
             )
             .owns(