Merge pull request #24906 from teskje/operator-hydration-logging

Compute operator hydration status logging

doc/user/content/sql/system-catalog/mz_internal.md

@@ -341,6 +341,22 @@ A compute object is hydrated on a given replica when it has fully processed the
 | `replica_id` | [`text`]    | The ID of a cluster replica. |
 | `hydrated`   | [`boolean`] | Whether the compute object is hydrated on the replica. |
 
+### `mz_compute_operator_hydration_statuses`
+
+The `mz_compute_operator_hydration_statuses` table describes the dataflow operator hydration status of compute objects (indexes or materialized views).
+
+A dataflow operator is hydrated on a given replica when it has fully processed the initial snapshot of data available in its inputs.
+
+<!-- RELATION_SPEC mz_internal.mz_compute_operator_hydration_statuses -->
+| Field                   | Type        | Meaning  |
+| ----------------------- | ----------- | -------- |
+| `object_id`             | [`text`]    | The ID of a compute object. Corresponds to [`mz_catalog.mz_indexes.id`](../mz_catalog#mz_indexes) or [`mz_catalog.mz_materialized_views.id`](../mz_catalog#mz_materialized_views). |
+| `physical_plan_node_id` | [`uint8`]   | The ID of a node in the physical plan of the compute object. Corresponds to a `node_id` displayed in the output of `EXPLAIN PHYSICAL PLAN WITH (node_ids)`. |
+| `replica_id`            | [`text`]    | The ID of a cluster replica. |
+| `hydrated`              | [`boolean`] | Whether the node is hydrated on the replica. |
+
+<!-- RELATION_SPEC_UNDOCUMENTED mz_internal.mz_compute_operator_hydration_statuses_per_worker -->
+
 ### `mz_frontiers`
 
 The `mz_frontiers` table describes the frontiers of each source, sink, table,

src/adapter/src/flags.rs

@@ -39,6 +39,9 @@ pub fn compute_config(config: &SystemVars) -> ComputeParameters {
         enable_mz_join_core: Some(config.enable_mz_join_core()),
         enable_jemalloc_profiling: Some(config.enable_jemalloc_profiling()),
         enable_columnation_lgalloc: Some(config.enable_columnation_lgalloc()),
+        enable_operator_hydration_status_logging: Some(
+            config.enable_compute_operator_hydration_status_logging(),
+        ),
         persist: persist_config(config),
         tracing: tracing_config(config),
         grpc_client: grpc_client_config(config),

src/catalog/src/builtin.rs

@@ -1946,6 +1946,20 @@ pub static MZ_COMPUTE_HYDRATION_STATUSES: Lazy<BuiltinSource> = Lazy::new(|| Bui
     is_retained_metrics_object: false,
     access: vec![PUBLIC_SELECT],
 });
+pub static MZ_COMPUTE_OPERATOR_HYDRATION_STATUSES_PER_WORKER: Lazy<BuiltinSource> =
+    Lazy::new(|| BuiltinSource {
+        name: "mz_compute_operator_hydration_statuses_per_worker",
+        schema: MZ_INTERNAL_SCHEMA,
+        data_source: IntrospectionType::ComputeOperatorHydrationStatus,
+        desc: RelationDesc::empty()
+            .with_column("object_id", ScalarType::String.nullable(false))
+            .with_column("physical_plan_node_id", ScalarType::UInt64.nullable(false))
+            .with_column("replica_id", ScalarType::String.nullable(false))
+            .with_column("worker_id", ScalarType::UInt64.nullable(false))
+            .with_column("hydrated", ScalarType::Bool.nullable(false)),
+        is_retained_metrics_object: false,
+        access: vec![PUBLIC_SELECT],
+    });
 
 pub static MZ_DATABASES: Lazy<BuiltinTable> = Lazy::new(|| BuiltinTable {
     name: "mz_databases",
@@ -4178,6 +4192,21 @@ HAVING pg_catalog.sum(count) != 0",
     access: vec![PUBLIC_SELECT],
 });
 
+pub static MZ_COMPUTE_OPERATOR_HYDRATION_STATUSES: Lazy<BuiltinView> = Lazy::new(|| BuiltinView {
+    name: "mz_compute_operator_hydration_statuses",
+    schema: MZ_INTERNAL_SCHEMA,
+    column_defs: None,
+    sql: "
+SELECT
+    object_id,
+    physical_plan_node_id,
+    replica_id,
+    bool_and(hydrated) AS hydrated
+FROM mz_internal.mz_compute_operator_hydration_statuses_per_worker
+GROUP BY object_id, physical_plan_node_id, replica_id",
+    access: vec![PUBLIC_SELECT],
+});
+
 pub static MZ_MESSAGE_COUNTS_PER_WORKER: Lazy<BuiltinView> = Lazy::new(|| BuiltinView {
     name: "mz_message_counts_per_worker",
     schema: MZ_INTERNAL_SCHEMA,
@@ -6711,10 +6740,12 @@ pub static BUILTINS_STATIC: Lazy<Vec<Builtin<NameReference>>> = Lazy::new(|| {
         Builtin::View(&MZ_GLOBAL_FRONTIERS),
         Builtin::Source(&MZ_COMPUTE_DEPENDENCIES),
         Builtin::Source(&MZ_COMPUTE_HYDRATION_STATUSES),
+        Builtin::Source(&MZ_COMPUTE_OPERATOR_HYDRATION_STATUSES_PER_WORKER),
         Builtin::View(&MZ_HYDRATION_STATUSES),
         Builtin::View(&MZ_MATERIALIZATION_LAG),
         Builtin::View(&MZ_COMPUTE_ERROR_COUNTS_PER_WORKER),
         Builtin::View(&MZ_COMPUTE_ERROR_COUNTS),
+        Builtin::View(&MZ_COMPUTE_OPERATOR_HYDRATION_STATUSES),
         Builtin::Source(&MZ_CLUSTER_REPLICA_FRONTIERS),
         Builtin::Source(&MZ_CLUSTER_REPLICA_HEARTBEATS),
         Builtin::Index(&MZ_SHOW_DATABASES_IND),

src/compute-client/src/controller.rs

@@ -651,9 +651,11 @@ where
         }
 
         for (type_, updates) in updates_by_type {
-            self.storage
-                .record_introspection_updates(type_, updates)
-                .await;
+            if !updates.is_empty() {
+                self.storage
+                    .record_introspection_updates(type_, updates)
+                    .await;
+            }
         }
     }
 }
@@ -704,13 +706,17 @@ where
 
     #[tracing::instrument(level = "debug", skip(self))]
     async fn maintain(&mut self) {
-        // Record pending introspection updates.
-        self.record_introspection_updates().await;
-
         // Perform instance maintenance work.
         for instance in self.compute.instances.values_mut() {
             instance.activate(self.storage).maintain();
         }
+
+        // Record pending introspection updates.
+        //
+        // It's beneficial to do this as the last maintenance step because previous steps can cause
+        // dropping of state, which can can cause introspection retractions, which lower the volume
+        // of data we have to record.
+        self.record_introspection_updates().await;
     }
 }
 

src/compute-client/src/controller/instance.rs

@@ -20,6 +20,7 @@ use futures::{future, StreamExt};
 use mz_build_info::BuildInfo;
 use mz_cluster_client::client::{ClusterStartupEpoch, TimelyConfig};
 use mz_compute_types::dataflows::DataflowDescription;
+use mz_compute_types::plan::NodeId;
 use mz_compute_types::sinks::{ComputeSinkConnection, ComputeSinkDesc, PersistSinkConnection};
 use mz_compute_types::sources::SourceInstanceDesc;
 use mz_expr::RowSetFinishing;
@@ -46,7 +47,8 @@ use crate::protocol::command::{
 };
 use crate::protocol::history::ComputeCommandHistory;
 use crate::protocol::response::{
-    ComputeResponse, CopyToResponse, PeekResponse, SubscribeBatch, SubscribeResponse,
+    ComputeResponse, CopyToResponse, OperatorHydrationStatus, PeekResponse, StatusResponse,
+    SubscribeBatch, SubscribeResponse,
 };
 use crate::service::{ComputeClient, ComputeGrpcClient};
 
@@ -444,6 +446,34 @@ impl<T: Timestamp> Instance<T> {
         }
     }
 
+    /// Update the tracked hydration status for an operator according to a received status update.
+    fn update_operator_hydration_status(
+        &mut self,
+        replica_id: ReplicaId,
+        status: OperatorHydrationStatus,
+    ) {
+        let Some(replica) = self.replicas.get_mut(&replica_id) else {
+            tracing::error!(
+                %replica_id, ?status,
+                "status update for an unknown replica"
+            );
+            return;
+        };
+        let Some(collection) = replica.collections.get_mut(&status.collection_id) else {
+            tracing::error!(
+                %replica_id, ?status,
+                "status update for an unknown collection"
+            );
+            return;
+        };
+
+        collection.hydration_state.operator_hydrated(
+            status.lir_id,
+            status.worker_id,
+            status.hydrated,
+        );
+    }
+
     /// Clean up collection state that is not needed anymore.
     ///
     /// Three conditions need to be true before we can remove state for a collection:
@@ -1500,6 +1530,10 @@ where
             ComputeResponse::SubscribeResponse(id, response) => {
                 self.handle_subscribe_response(id, response, replica_id)
             }
+            ComputeResponse::Status(response) => {
+                self.handle_status_response(response, replica_id);
+                None
+            }
         }
     }
 
@@ -1689,6 +1723,14 @@ where
             }
         }
     }
+
+    fn handle_status_response(&mut self, response: StatusResponse, replica_id: ReplicaId) {
+        match response {
+            StatusResponse::OperatorHydration(status) => self
+                .compute
+                .update_operator_hydration_status(replica_id, status),
+        }
+    }
 }
 
 impl<'a, T> ActiveInstance<'a, T>
@@ -1854,12 +1896,12 @@ impl<T> ReplicaState<T> {
     /// Add a collection to the replica state.
     fn add_collection(&mut self, id: GlobalId, as_of: Antichain<T>) {
         let metrics = self.metrics.for_collection(id);
-        let hydration_flag = HydrationFlag::new(self.id, id, self.introspection_tx.clone());
+        let hydration_state = HydrationState::new(self.id, id, self.introspection_tx.clone());
         let mut state = ReplicaCollectionState {
             metrics,
             created_at: Instant::now(),
             as_of,
-            hydration_flag,
+            hydration_state,
         };
 
         // We need to consider the edge case where the as-of is the empty frontier. Such an as-of
@@ -1897,14 +1939,14 @@ struct ReplicaCollectionState<T> {
     created_at: Instant,
     /// As-of frontier with which this collection was installed on the replica.
     as_of: Antichain<T>,
-    /// Tracks whether this collection is hydrated, i.e., it has produced some initial output.
-    hydration_flag: HydrationFlag,
+    /// Tracks hydration state for this collection.
+    hydration_state: HydrationState,
 }
 
 impl<T> ReplicaCollectionState<T> {
     /// Returns whether this collection is hydrated.
     fn hydrated(&self) -> bool {
-        self.hydration_flag.hydrated
+        self.hydration_state.hydrated
     }
 
     /// Marks the collection as hydrated and updates metrics and introspection accordingly.
@@ -1914,22 +1956,29 @@ impl<T> ReplicaCollectionState<T> {
             metrics.initial_output_duration_seconds.set(duration);
         }
 
-        self.hydration_flag.set();
+        self.hydration_state.collection_hydrated();
     }
 }
 
-/// A wrapper type that maintains hydration introspection for a given replica and collection, and
-/// ensures that reported introspection data is retracted when the flag is dropped.
+/// Maintains both global and operator-level hydration introspection for a given replica and
+/// collection, and ensures that reported introspection data is retracted when the flag is dropped.
 #[derive(Debug)]
-struct HydrationFlag {
+struct HydrationState {
+    /// The ID of the replica.
     replica_id: ReplicaId,
+    /// The ID of the compute collection.
     collection_id: GlobalId,
+    /// Whether the collection is hydrated.
     hydrated: bool,
+    /// Operator-level hydration state.
+    /// (lir_id, worker_id) -> hydrated
+    operators: BTreeMap<(NodeId, usize), bool>,
+    /// A channel through which introspection updates are delivered.
     introspection_tx: crossbeam_channel::Sender<IntrospectionUpdates>,
 }
 
-impl HydrationFlag {
-    /// Create a new unset `HydrationFlag` and update introspection.
+impl HydrationState {
+    /// Create a new `HydrationState` and initialize introspection.
     fn new(
         replica_id: ReplicaId,
         collection_id: GlobalId,
@@ -1939,55 +1988,110 @@ impl HydrationFlag {
             replica_id,
             collection_id,
             hydrated: false,
+            operators: Default::default(),
             introspection_tx,
         };
 
-        let insertion = self_.row();
-        self_.send(vec![(insertion, 1)]);
+        let insertion = self_.row_for_collection();
+        self_.send(
+            IntrospectionType::ComputeHydrationStatus,
+            vec![(insertion, 1)],
+        );
 
         self_
     }
 
-    /// Mark the collection as hydrated and update introspection.
-    fn set(&mut self) {
+    /// Update the collection as hydrated.
+    fn collection_hydrated(&mut self) {
         if self.hydrated {
             return; // nothing to do
         }
 
-        let retraction = self.row();
+        let retraction = self.row_for_collection();
         self.hydrated = true;
-        let insertion = self.row();
+        let insertion = self.row_for_collection();
+
+        self.send(
+            IntrospectionType::ComputeHydrationStatus,
+            vec![(retraction, -1), (insertion, 1)],
+        );
+    }
+
+    /// Update the given (lir_id, worker_id) pair as hydrated.
+    fn operator_hydrated(&mut self, lir_id: NodeId, worker_id: usize, hydrated: bool) {
+        let retraction = self.row_for_operator(lir_id, worker_id);
+        self.operators.insert((lir_id, worker_id), hydrated);
+        let insertion = self.row_for_operator(lir_id, worker_id);
+
+        if retraction == insertion {
+            return; // no change
+        }
 
-        self.send(vec![(retraction, -1), (insertion, 1)]);
+        let updates = retraction
+            .map(|r| (r, -1))
+            .into_iter()
+            .chain(insertion.map(|r| (r, 1)))
+            .collect();
+        self.send(IntrospectionType::ComputeOperatorHydrationStatus, updates);
     }
 
-    fn row(&self) -> Row {
+    /// Return a `Row` reflecting the current collection hydration status.
+    fn row_for_collection(&self) -> Row {
         Row::pack_slice(&[
             Datum::String(&self.collection_id.to_string()),
             Datum::String(&self.replica_id.to_string()),
             Datum::from(self.hydrated),
         ])
     }
 
-    fn send(&self, updates: Vec<(Row, Diff)>) {
-        let result = self
-            .introspection_tx
-            .send((IntrospectionType::ComputeHydrationStatus, updates));
+    /// Return a `Row` reflecting the current hydration status of the identified operator.
+    ///
+    /// Returns `None` if the identified operator is not tracked.
+    fn row_for_operator(&self, lir_id: NodeId, worker_id: usize) -> Option<Row> {
+        self.operators.get(&(lir_id, worker_id)).map(|hydrated| {
+            Row::pack_slice(&[
+                Datum::String(&self.collection_id.to_string()),
+                Datum::UInt64(lir_id),
+                Datum::String(&self.replica_id.to_string()),
+                Datum::UInt64(u64::cast_from(worker_id)),
+                Datum::from(*hydrated),
+            ])
+        })
+    }
+
+    fn send(&self, introspection_type: IntrospectionType, updates: Vec<(Row, Diff)>) {
+        let result = self.introspection_tx.send((introspection_type, updates));
 
         if result.is_err() {
             // The global controller holds on to the `introspection_rx`. So when we get here that
             // probably means that the controller was dropped and the process is shutting down, in
             // which case we don't care about introspection updates anymore.
             tracing::info!(
-                "discarding `ComputeHydrationStatus` update because the receiver disconnected"
+                ?introspection_type,
+                "discarding introspection update because the receiver disconnected"
             );
         }
     }
 }
 
-impl Drop for HydrationFlag {
+impl Drop for HydrationState {
     fn drop(&mut self) {
-        let retraction = self.row();
-        self.send(vec![(retraction, -1)]);
+        // Retract collection hydration status.
+        let retraction = self.row_for_collection();
+        self.send(
+            IntrospectionType::ComputeHydrationStatus,
+            vec![(retraction, -1)],
+        );
+
+        // Retract operator-level hydration status.
+        let operators: Vec<_> = self.operators.keys().collect();
+        let updates: Vec<_> = operators
+            .into_iter()
+            .flat_map(|(lir_id, worker_id)| self.row_for_operator(*lir_id, *worker_id))
+            .map(|r| (r, -1))
+            .collect();
+        if !updates.is_empty() {
+            self.send(IntrospectionType::ComputeOperatorHydrationStatus, updates)
+        }
     }
 }