Skip to content

fix: ensure cache affinity in Mod partition shuffle #19309

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
dantengsky wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

fix: ensure cache affinity in Mod partition shuffle #19309

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
29 changes: 21 additions & 8 deletions src/query/catalog/src/plan/partition.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -136,14 +136,27 @@ impl Partitions {
let partitions = match self.kind {
PartitionsShuffleKind::Seq => self.partitions.clone(),
PartitionsShuffleKind::Mod => {
// Sort by hash%executor_nums.
let mut parts = self
.partitions
.iter()
.map(|p| (p.hash() % num_executors as u64, p.clone()))
.collect::<Vec<_>>();
parts.sort_by(|a, b| a.0.cmp(&b.0));
parts.into_iter().map(|x| x.1).collect()
// Assign partitions to executors by hash % num_executors.
// This ensures cache affinity: the same partition always goes to the same executor,
// as long as the cluster membership remains unchanged.
let mut executor_part: HashMap<String, Partitions> = HashMap::new();

for p in self.partitions.iter() {
let idx = (p.hash() % num_executors as u64) as usize;
let executor_id = &executors_sorted[idx].id;
executor_part
.entry(executor_id.clone())
.or_default()
.partitions
.push(p.clone());
}

// Ensure all executors have an entry (even if empty).
for e in &executors_sorted {
executor_part.entry(e.id.clone()).or_default();
}

return Ok(executor_part);
}
PartitionsShuffleKind::ConsistentHash => {
let mut scale = 0;
Expand Down
16 changes: 8 additions & 8 deletions src/query/catalog/tests/it/testdata/partition-reshuffle.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,16 +7,16 @@ Partitions { kind: Seq, partitions: [] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"0"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"1"}] }
PartitionsShuffleKind::Mod : 10 partitions of 3 executors
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"0"}, {"type":"fuse_lazy","loc":"2"}, {"type":"fuse_lazy","loc":"4"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"9"}, {"type":"fuse_lazy","loc":"1"}, {"type":"fuse_lazy","loc":"3"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"5"}, {"type":"fuse_lazy","loc":"7"}, {"type":"fuse_lazy","loc":"6"}, {"type":"fuse_lazy","loc":"8"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"0"}, {"type":"fuse_lazy","loc":"2"}, {"type":"fuse_lazy","loc":"4"}, {"type":"fuse_lazy","loc":"9"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"1"}, {"type":"fuse_lazy","loc":"3"}, {"type":"fuse_lazy","loc":"5"}, {"type":"fuse_lazy","loc":"7"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"6"}, {"type":"fuse_lazy","loc":"8"}] }
PartitionsShuffleKind::Mod : 11 partitions of 3 executors
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"0"}, {"type":"fuse_lazy","loc":"2"}, {"type":"fuse_lazy","loc":"4"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"9"}, {"type":"fuse_lazy","loc":"1"}, {"type":"fuse_lazy","loc":"3"}, {"type":"fuse_lazy","loc":"5"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"7"}, {"type":"fuse_lazy","loc":"10"}, {"type":"fuse_lazy","loc":"6"}, {"type":"fuse_lazy","loc":"8"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"0"}, {"type":"fuse_lazy","loc":"2"}, {"type":"fuse_lazy","loc":"4"}, {"type":"fuse_lazy","loc":"9"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"1"}, {"type":"fuse_lazy","loc":"3"}, {"type":"fuse_lazy","loc":"5"}, {"type":"fuse_lazy","loc":"7"}, {"type":"fuse_lazy","loc":"10"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"6"}, {"type":"fuse_lazy","loc":"8"}] }
PartitionsShuffleKind::Mod : 11 partitions of 2 executors
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"4"}, {"type":"fuse_lazy","loc":"5"}, {"type":"fuse_lazy","loc":"0"}, {"type":"fuse_lazy","loc":"1"}, {"type":"fuse_lazy","loc":"2"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"3"}, {"type":"fuse_lazy","loc":"6"}, {"type":"fuse_lazy","loc":"7"}, {"type":"fuse_lazy","loc":"8"}, {"type":"fuse_lazy","loc":"9"}, {"type":"fuse_lazy","loc":"10"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"4"}, {"type":"fuse_lazy","loc":"5"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"0"}, {"type":"fuse_lazy","loc":"1"}, {"type":"fuse_lazy","loc":"2"}, {"type":"fuse_lazy","loc":"3"}, {"type":"fuse_lazy","loc":"6"}, {"type":"fuse_lazy","loc":"7"}, {"type":"fuse_lazy","loc":"8"}, {"type":"fuse_lazy","loc":"9"}, {"type":"fuse_lazy","loc":"10"}] }
PartitionsShuffleKind::ConsistentHash : 11 partitions of 2 executors
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"4"}, {"type":"fuse_lazy","loc":"5"}, {"type":"fuse_lazy","loc":"6"}, {"type":"fuse_lazy","loc":"7"}] }
Partitions { kind: Seq, partitions: [{"type":"fuse_lazy","loc":"0"}, {"type":"fuse_lazy","loc":"1"}, {"type":"fuse_lazy","loc":"2"}, {"type":"fuse_lazy","loc":"3"}, {"type":"fuse_lazy","loc":"8"}, {"type":"fuse_lazy","loc":"9"}, {"type":"fuse_lazy","loc":"10"}] }
Expand Down
Loading