fix: add table status to cancel background jobs

Cargo.toml

@@ -94,6 +94,7 @@ analytic_engine = { path = "analytic_engine" }
 arena = { path = "components/arena" }
 async-stream = "0.3.4"
 async-trait = "0.1.72"
+atomic_enum = "0.2.0"
 base64 = "0.13"
 bytes = "1"
 bytes_ext = { path = "components/bytes_ext" }

analytic_engine/Cargo.toml

@@ -34,6 +34,7 @@ arena = { workspace = true }
 arrow = { workspace = true }
 async-stream = { workspace = true }
 async-trait = { workspace = true }
+atomic_enum = { workspace = true }
 base64 = { workspace = true }
 bytes_ext = { workspace = true }
 ceresdbproto = { workspace = true }

analytic_engine/src/compaction/scheduler.rs

@@ -569,6 +569,14 @@ impl ScheduleWorker {
 
     async fn handle_table_compaction_request(&self, compact_req: TableCompactionRequest) {
         let table_data = compact_req.table_data.clone();
+        if !table_data.allow_compaction() {
+            error!(
+                "Table status is not ok, unable to compact further, table:{}, table_id:{}",
+                table_data.name, table_data.id
+            );
+            return;
+        }
+
         let table_options = table_data.table_options();
         let compaction_strategy = table_options.compaction_strategy;
         let picker = self.picker_manager.get_picker(compaction_strategy);

analytic_engine/src/instance/close.rs

@@ -79,6 +79,10 @@ impl Closer {
         let removed_table = self.space.remove_table(&request.table_name);
         assert!(removed_table.is_some());
 
+        // Table is already moved out of space, we should close it to stop background
+        // jobs.
+        table_data.set_closed();
+
         info!(
             "table:{}-{} has been removed from the space_id:{}",
             table_data.name, table_data.id, self.space.id

analytic_engine/src/instance/flush_compaction.rs

@@ -725,6 +725,16 @@ impl SpaceStore {
             .await?;
         }
 
+        if !table_data.allow_compaction() {
+            return Other {
+                msg: format!(
+                    "Table status is not ok, unable to update manifest, table:{}, table_id:{}",
+                    table_data.name, table_data.id
+                ),
+            }
+            .fail();
+        }
+
         let edit_req = {
             let meta_update = MetaUpdate::VersionEdit(edit_meta.clone());
             MetaEditRequest {

analytic_engine/src/table/data.rs

@@ -21,7 +21,7 @@ use std::{
     fmt::Formatter,
     num::NonZeroUsize,
     sync::{
-        atomic::{AtomicBool, AtomicU32, AtomicU64, AtomicUsize, Ordering},
+        atomic::{AtomicU32, AtomicU64, AtomicUsize, Ordering},
         Arc, Mutex,
     },
     time::Duration,
@@ -111,6 +111,28 @@ impl TableShardInfo {
     }
 }
 
+/// `atomic_enum` macro will expand method like
+/// ```text
+/// compare_exchange(..) -> Result<TableStatus, TableStatus>
+/// ```
+/// The result type is conflict with outer
+/// Result, so add this hack
+// TODO: fix this in atomic_enum crate.
+mod hack {
+    use atomic_enum::atomic_enum;
+
+    #[atomic_enum]
+    #[derive(PartialEq)]
+    pub enum TableStatus {
+        Ok = 0,
+        Closed,
+        /// No write/alter are allowed after table is dropped.
+        Dropped,
+    }
+}
+
+use self::hack::{AtomicTableStatus, TableStatus};
+
 /// Data of a table
 pub struct TableData {
     /// Id of this table
@@ -161,10 +183,8 @@ pub struct TableData {
     /// Not persist, used to determine if this table should flush.
     last_flush_time_ms: AtomicU64,
 
-    /// Flag denoting whether the table is dropped
-    ///
-    /// No write/alter is allowed if the table is dropped.
-    dropped: AtomicBool,
+    /// Table Status
+    status: AtomicTableStatus,
 
     /// Manifest updates after last snapshot
     manifest_updates: AtomicUsize,
@@ -192,7 +212,7 @@ impl fmt::Debug for TableData {
             .field("opts", &self.opts)
             .field("last_sequence", &self.last_sequence)
             .field("last_memtable_id", &self.last_memtable_id)
-            .field("dropped", &self.dropped.load(Ordering::Relaxed))
+            .field("status", &self.status.load(Ordering::Relaxed))
             .field("shard_info", &self.shard_info)
             .finish()
     }
@@ -265,7 +285,7 @@ impl TableData {
             last_memtable_id: AtomicU64::new(0),
             allocator: IdAllocator::new(0, 0, DEFAULT_ALLOC_STEP),
             last_flush_time_ms: AtomicU64::new(0),
-            dropped: AtomicBool::new(false),
+            status: TableStatus::Ok.into(),
             metrics,
             shard_info: TableShardInfo::new(shard_id),
             serial_exec: tokio::sync::Mutex::new(TableOpSerialExecutor::new(table_id)),
@@ -310,7 +330,7 @@ impl TableData {
             last_memtable_id: AtomicU64::new(0),
             allocator,
             last_flush_time_ms: AtomicU64::new(0),
-            dropped: AtomicBool::new(false),
+            status: TableStatus::Ok.into(),
             metrics,
             shard_info: TableShardInfo::new(shard_id),
             serial_exec: tokio::sync::Mutex::new(TableOpSerialExecutor::new(add_meta.table_id)),
@@ -382,13 +402,26 @@ impl TableData {
 
     #[inline]
     pub fn is_dropped(&self) -> bool {
-        self.dropped.load(Ordering::SeqCst)
+        self.status.load(Ordering::SeqCst) == TableStatus::Dropped
     }
 
     /// Set the table is dropped and forbid any writes/alter on this table.
     #[inline]
     pub fn set_dropped(&self) {
-        self.dropped.store(true, Ordering::SeqCst);
+        self.status.store(TableStatus::Dropped, Ordering::SeqCst)
+    }
+
+    #[inline]
+    pub fn set_closed(&self) {
+        self.status.store(TableStatus::Closed, Ordering::SeqCst)
+    }
+
+    #[inline]
+    pub fn allow_compaction(&self) -> bool {
+        match self.status.load(Ordering::SeqCst) {
+            TableStatus::Ok => true,
+            TableStatus::Closed | TableStatus::Dropped => false,
+        }
     }
 
     /// Returns total memtable memory usage in bytes.