pageserver: add GcInfo::notify_child_keyspace

pageserver/src/tenant.rs

@@ -19,6 +19,7 @@ use enumset::EnumSet;
 use futures::stream::FuturesUnordered;
 use futures::FutureExt;
 use futures::StreamExt;
+use pageserver_api::keyspace::KeySpace;
 use pageserver_api::models;
 use pageserver_api::models::AuxFilePolicy;
 use pageserver_api::models::TimelineState;
@@ -2786,11 +2787,12 @@ impl Tenant {
 
         // Scan all timelines. For each timeline, remember the timeline ID and
         // the branch point where it was created.
-        let mut all_branchpoints: BTreeMap<TimelineId, Vec<(Lsn, TimelineId)>> = BTreeMap::new();
+        let mut all_branchpoints: BTreeMap<TimelineId, Vec<(Lsn, TimelineId, Option<KeySpace>)>> =
+            BTreeMap::new();
         timelines.iter().for_each(|(timeline_id, timeline_entry)| {
             if let Some(ancestor_timeline_id) = &timeline_entry.get_ancestor_timeline_id() {
                 let ancestor_children = all_branchpoints.entry(*ancestor_timeline_id).or_default();
-                ancestor_children.push((timeline_entry.get_ancestor_lsn(), *timeline_id));
+                ancestor_children.push((timeline_entry.get_ancestor_lsn(), *timeline_id, None));
             }
         });
 
@@ -2799,7 +2801,7 @@ impl Tenant {
 
         // Populate each timeline's GcInfo with information about its child branches
         for timeline in timelines.values() {
-            let mut branchpoints: Vec<(Lsn, TimelineId)> = all_branchpoints
+            let mut branchpoints: Vec<(Lsn, TimelineId, Option<KeySpace>)> = all_branchpoints
                 .remove(&timeline.timeline_id)
                 .unwrap_or_default();
 
@@ -4305,7 +4307,7 @@ mod tests {
         {
             let branchpoints = &tline.gc_info.read().unwrap().retain_lsns;
             assert_eq!(branchpoints.len(), 1);
-            assert_eq!(branchpoints[0], (Lsn(0x40), NEW_TIMELINE_ID));
+            assert_eq!(branchpoints[0], (Lsn(0x40), NEW_TIMELINE_ID, None));
         }
 
         // You can read the key from the child branch even though the parent is

pageserver/src/tenant/size.rs

@@ -271,10 +271,14 @@ pub(super) async fn gather_inputs(
         let mut lsns: Vec<(Lsn, LsnKind)> = gc_info
             .retain_lsns
             .iter()
-            .filter(|(lsn, _child_id)| lsn > &ancestor_lsn)
-            .copied()
-            // this assumes there are no other retain_lsns than the branchpoints
-            .map(|(lsn, _child_id)| (lsn, LsnKind::BranchPoint))
+            .filter_map(|(lsn, _child_id, _)| {
+                if lsn > &ancestor_lsn {
+                    // this assumes there are no other retain_lsns than the branchpoints
+                    Some((*lsn, LsnKind::BranchPoint))
+                } else {
+                    None
+                }
+            })
             .collect::<Vec<_>>();
 
         lsns.extend(lease_points.iter().map(|&lsn| (lsn, LsnKind::LeasePoint)));

pageserver/src/tenant/timeline.rs

@@ -453,12 +453,12 @@ pub struct WalReceiverInfo {
 /// Garbage Collection.
 #[derive(Default)]
 pub(crate) struct GcInfo {
-    /// Specific LSNs that are needed.
+    /// Record which parts of this timeline's history are still needed by children
     ///
-    /// Currently, this includes all points where child branches have
-    /// been forked off from. In the future, could also include
-    /// explicit user-defined snapshot points.
-    pub(crate) retain_lsns: Vec<(Lsn, TimelineId)>,
+    /// Optionally store each child's keyspace at their branch LSN: parts of the keyspace not covered here may be dropped during GC, as
+    /// the child will never read them.  For example, a child which has covered its whole keyspace with image layers
+    /// will put an empty keyspace here.  Children populate this: if it is None, presume the child may read any part of the keyspace.
+    pub(crate) retain_lsns: Vec<(Lsn, TimelineId, Option<KeySpace>)>,
 
     /// The cutoff coordinates, which are combined by selecting the minimum.
     pub(crate) cutoffs: GcCutoffs,
@@ -476,13 +476,21 @@ impl GcInfo {
     }
 
     pub(super) fn insert_child(&mut self, child_id: TimelineId, child_lsn: Lsn) {
-        self.retain_lsns.push((child_lsn, child_id));
+        self.retain_lsns.push((child_lsn, child_id, None));
         self.retain_lsns.sort_by_key(|i| i.0);
     }
 
     pub(super) fn remove_child(&mut self, child_id: TimelineId) {
         self.retain_lsns.retain(|i| i.1 != child_id);
     }
+
+    /// When the child re-calculates which parts of the keyspace it will read from the ancestor, it posts
+    /// and update to the parent using this function, to enable the parent to perhaps GC more layers.
+    pub(super) fn notify_child_keyspace(&mut self, child_id: TimelineId, key_space: KeySpace) {
+        if let Ok(idx) = self.retain_lsns.binary_search_by_key(&child_id, |i| i.1) {
+            self.retain_lsns.get_mut(idx).unwrap().2 = Some(key_space);
+        }
+    }
 }
 
 /// The `GcInfo` component describing which Lsns need to be retained.
@@ -5089,7 +5097,7 @@ impl Timeline {
             let retain_lsns = gc_info
                 .retain_lsns
                 .iter()
-                .map(|(lsn, _child_id)| *lsn)
+                .map(|(lsn, _child_id, _)| *lsn)
                 .collect();
 
             // Gets the maximum LSN that holds the valid lease.