wip

Author: dantengsky

Cargo.lock

@@ -82,7 +82,7 @@ join_spilling_memory_ratio = 60
 [log]
 
 [log.file]
-level = "INFO"
+level = "DEBUG"
 format = "text"
 dir = "./.databend/logs_1"
 limit = 12 # 12 files, 1 file per hour

scripts/ci/deploy/config/databend-query-node-1.toml

@@ -92,13 +92,15 @@ pub async fn do_vacuum2(
 
     let retention_policy = fuse_table.get_data_retention_policy(ctx.as_ref())?;
 
+    // Indicates whether to use the current table snapshot as gc root,
+    // true means vacuum all the historical snapshots.
     let mut is_vacuum_all = false;
 
     let mut respect_flash_back_with_lvt = None;
     let snapshots_before_lvt = match retention_policy {
         RetentionPolicy::ByTimePeriod(delta_duration) => {
             let retention_period = if fuse_table.is_transient() {
-                // for transient table, keep no history data
+                // For transient table, keep no history data
                 TimeDelta::zero()
             } else {
                 delta_duration
@@ -176,15 +178,14 @@ pub async fn do_vacuum2(
             }
             if num_snapshot_to_keep == 1 {
                 // Expecting only one snapshot left, which means that we can use the current snapshot
-                // as gc root, this flag will be propagated to the select_gc_root routine later.
+                // as gc root, this flag will be propagated to the select_gc_root func later.
                 is_vacuum_all = true;
             }
 
-            // Later while selecting gc root, that last one of the `snapshots` will be used as gc
-            // candidate(which we are not sure about if it has been successfully committed).
+            // When selecting the GC root later, the last snapshot in `snapshots` is a candidate,
+            // but its commit status is uncertain, its previous snapshot is typically used as the GC root, except in the is_vacuum_all case.
             //
-            // The previous snapshot of the candidate will be used as the GC root (expect for the is_vacuum_all scenario).
-            // Thus while truncating the snapshots, we should keep 2 extras snapshots; see `select_gc_root` for more details.
+            // Therefore, during snapshot truncation, we keep 2 extra snapshots; see `select_gc_root` for details.
             let num_candidates = len - num_snapshot_to_keep + 2;
             snapshots.truncate(num_candidates);
             snapshots

src/query/ee/src/storages/fuse/operations/vacuum_table_v2.rs

@@ -10,7 +10,6 @@ edition = { workspace = true }
 [dependencies]
 async-backtrace = { workspace = true }
 async-trait = { workspace = true }
-chrono = { workspace = true }
 databend-common-base = { workspace = true }
 databend-common-catalog = { workspace = true }
 databend-common-exception = { workspace = true }