Fix slot-range parsing during migration and add rebalance planning logic#92
Merged
Merged
Conversation
ysqyang
force-pushed
the
cluster-state-classification
branch
from
34325d0 to
78cb392
Compare
ysqyang
changed the title
ysqyang
force-pushed
the
cluster-state-classification
branch
4 times, most recently
from
ede2f24 to
e728115
Compare
ysqyang
changed the title
ysqyang
force-pushed
the
cluster-state-classification
branch
4 times, most recently
from
170677e to
ea75281
Compare
jdheyburn
reviewed
Feb 26, 2026
Collaborator
jdheyburn
left a comment
jdheyburn
left a comment
There was a problem hiding this comment.
Looks good, though I know nothing about rebalancing. I feel having clearer variable names would make it easier for a non-Cluster person like me to understand it more, which is where my comments are around.
If there are opportunities to extract a block of code into a function behind a descriptive name, that would be great too!
bjosv
reviewed
Feb 26, 2026
ysqyang
force-pushed
the
cluster-state-classification
branch
4 times, most recently
from
6621f41 to
8cb1309
Compare
bjosv
reviewed
Feb 27, 2026
Collaborator
bjosv
left a comment
bjosv
left a comment
There was a problem hiding this comment.
Looks good, some comments.
Feels like the "Tolerate ±1-slot rounding differences" is a bit tricky.
|
|
||
| // BuildRebalanceMove computes a single, deterministic slot move to improve balance. | ||
| // It returns nil when the cluster is already balanced or not ready for rebalancing. | ||
| func BuildRebalanceMove(shards []*ShardState, expectedShards int, maxSlots int) (*SlotMove, error) { |
Collaborator
There was a problem hiding this comment.
We build a SlotMove in this function, is the name of this function misguiding?
Maybe its a PlanRebalanceMove?
Collaborator
There was a problem hiding this comment.
also, the shards []*ShardState, is that from the current ClusterState, or is it planned to be a subset?
Contributor
Author
There was a problem hiding this comment.
also, the
shards []*ShardState, is that from the current ClusterState, or is it planned to be a subset?
It's actually a superset of current ClusterState, which will become apparent in a follow-up PR.
| type SlotMove struct { | ||
| Src *NodeState | ||
| Dst *NodeState | ||
| Slots []int |
Collaborator
There was a problem hiding this comment.
Not for this PR; but maybe we should have a defined type for a Slot in clusterstate.go (or alias).
That might avoid any confusion if an int means a slot, a number of slots, a length or size.
ysqyang
force-pushed
the
cluster-state-classification
branch
2 times, most recently
from
724ec1f to
665422a
Compare
bjosv
approved these changes
Feb 27, 2026
Collaborator
bjosv
left a comment
bjosv
left a comment
There was a problem hiding this comment.
There are improvement possibilities, but I think we can address them in coming PRs.
ysqyang
force-pushed
the
cluster-state-classification
branch
3 times, most recently
from
760cab6 to
8766ab2
Compare
Two independent but related changes for scale-out support: 1. Skip migrating/importing entries (e.g. "[5461->-abc123]") in parseSlotsRanges to prevent parse errors when CLUSTER NODES output contains in-progress slot migrations. Adds unit tests. 2. Introduce PlanRebalanceMove which computes a single, deterministic slot move to incrementally rebalance a cluster after scale-out. Calculates per-primary slot targets (16384 / shards), identifies the most-loaded source and least-loaded destination, and returns a bounded SlotMove. Returns one move per call so each reconcile loop stays fast and restartable. Includes unit tests. Signed-off-by: yang.qiu <yang.qiu@reddit.com>
ysqyang
force-pushed
the
cluster-state-classification
branch
from
8766ab2 to
e094421
Compare
… logic - Rename rebalance.go → cluster_rebalance.go (and test file) - Rename BuildRebalanceMove → PlanRebalanceMove - Improve variable names for clarity (primarySlots fields, loop vars) - Move ±1 tolerance check into slot allocation loop to avoid masking real imbalances - Remove unused error return from takeSlotsFromRanges - Add remainder distribution comment in assignSlotsToPendingPrimaries Signed-off-by: yang.qiu <yang.qiu@reddit.com>
ysqyang
force-pushed
the
cluster-state-classification
branch
from
e094421 to
4d3314b
Compare
sandeepkunusoth
pushed a commit
to sandeepkunusoth/valkey-k8s-operator
that referenced
this pull request
Feb 28, 2026
…gic (valkey-io#92) - Fix `parseSlotsRanges` to skip `[slot->-nodeid]` / `[slot-<-nodeid]` entries that Valkey appends to `CLUSTER NODES` output during active slot migrations, which previously caused parse errors. - Add `BuildRebalanceMove` — a deterministic, incremental slot rebalance planner that computes a single slot migration from an overloaded primary to an underloaded (or empty) one. This is the planning layer for scale-out support; execution is wired in a follow-up PR. Signed-off-by: yang.qiu <yang.qiu@reddit.com> Co-authored-by: yang.qiu <yang.qiu@reddit.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
parseSlotsRangesto skip[slot->-nodeid]/[slot-<-nodeid]entries that Valkey appends toCLUSTER NODESoutput during active slot migrations, which previously caused parse errors.BuildRebalanceMove— a deterministic, incremental slot rebalance planner that computes a single slot migration from an overloaded primary to an underloaded (or empty) one. This is the planning layer for scale-out support; execution is wired in a follow-up PR.Test plan
parseSlotsRanges: normal ranges, migrating/importing entries, empty input, migration-only inputPlanRebalanceMove: scale-out (2→3 shards), balanced cluster (no-op), mismatched shard count, extra empty shards, zero max-slots, slot extraction from ranges