seaweedfs

13443 Commits

128 Branches

331 Tags

368 MiB

Tree: 4fe3df6b1d

Commit Graph

Author	SHA1	Message	Date
Chris Lu	5f85bf5e8a	Batch volume balance: run multiple moves per job (#8561 ) * proto: add BalanceMoveSpec and batch fields to BalanceTaskParams Add BalanceMoveSpec message for encoding individual volume moves, and max_concurrent_moves + repeated moves fields to BalanceTaskParams to support batching multiple volume moves in a single job. * balance handler: add batch execution with concurrent volume moves Refactor Execute() into executeSingleMove() (backward compatible) and executeBatchMoves() which runs multiple volume moves concurrently using a semaphore-bounded goroutine pool. When BalanceTaskParams.Moves is populated, the batch path is taken; otherwise the single-move path. Includes aggregate progress reporting across concurrent moves, per-move error collection, and partial failure support. * balance handler: add batch config fields to Descriptor and worker config Add max_concurrent_moves and batch_size fields to the worker config form and deriveBalanceWorkerConfig(). These control how many volume moves run concurrently within a batch job and the maximum batch size. * balance handler: group detection proposals into batch jobs When batch_size > 1, the Detect method groups detection results into batch proposals where each proposal encodes multiple BalanceMoveSpec entries in BalanceTaskParams.Moves. Single-result batches fall back to the existing single-move proposal format for backward compatibility. * admin UI: add volume balance execution plan and batch badge Add renderBalanceExecutionPlan() for rich rendering of volume balance jobs in the job detail modal. Single-move jobs show source/target/volume info; batch jobs show a moves table with all volume moves. Add batch badge (e.g., "5 moves") next to job type in the execution jobs table when the job has batch=true label. * Update plugin_templ.go * fix: detection algorithm uses greedy target instead of divergent topology scores The detection loop tracked effective volume counts via an adjustments map, but createBalanceTask independently called planBalanceDestination which used the topology's LoadCount — a separate, unadjusted source of truth. This divergence caused multiple moves to pile onto the same server. Changes: - Add resolveBalanceDestination to resolve the detection loop's greedy target (minServer) rather than independently picking a destination - Add oscillation guard: stop when max-min <= 1 since no single move can improve the balance beyond that point - Track unseeded destinations: if a target server wasn't in the initial serverVolumeCounts, add it so subsequent iterations include it - Add TestDetection_UnseededDestinationDoesNotOverload * fix: handler force_move propagation, partial failure, deterministic dedupe - Propagate ForceMove from outer BalanceTaskParams to individual move TaskParams so batch moves respect the force_move flag - Fix partial failure: mark job successful if at least one move succeeded (succeeded > 0 \|\| failed == 0) to avoid re-running already-completed moves on retry - Use SHA-256 hash for deterministic dedupe key fallback instead of time.Now().UnixNano() which is non-deterministic - Remove unused successDetails variable - Extract maxProposalStringLength constant to replace magic number 200 * admin UI: use template literals in balance execution plan rendering * fix: integration test handles batch proposals from batched detection With batch_size=20, all moves are grouped into a single proposal containing BalanceParams.Moves instead of top-level Sources/Targets. Update assertions to handle both batch and single-move proposal formats. * fix: verify volume size on target before deleting source during balance Add a pre-delete safety check that reads the volume file status on both source and target, then compares .dat file size and file count. If they don't match, the move is aborted — leaving the source intact rather than risking irreversible data loss. Also removes the redundant mountVolume call since VolumeCopy already mounts the volume on the target server. * fix: clamp maxConcurrent, serialize progress sends, validate config as int64 - Clamp maxConcurrentMoves to defaultMaxConcurrentMoves before creating the semaphore so a stale or malicious job cannot request unbounded concurrent volume moves - Extend progressMu to cover sender.SendProgress calls since the underlying gRPC stream is not safe for concurrent writes - Perform bounds checks on max_concurrent_moves and batch_size in int64 space before casting to int, avoiding potential overflow on 32-bit * fix: check disk capacity in resolveBalanceDestination Skip disks where VolumeCount >= MaxVolumeCount so the detection loop does not propose moves to a full disk that would fail at execution time. * test: rename unseeded destination test to match actual behavior The test exercises a server with 0 volumes that IS seeded from topology (matching disk type), not an unseeded destination. Rename to TestDetection_ZeroVolumeServerIncludedInBalance and fix comments. * test: tighten integration test to assert exactly one batch proposal With default batch_size=20, all moves should be grouped into a single batch proposal. Assert len(proposals)==1 and require BalanceParams with Moves, removing the legacy single-move else branch. * fix: propagate ctx to RPCs and restore source writability on abort - All helper methods (markVolumeReadonly, copyVolume, tailVolume, readVolumeFileStatus, deleteVolume) now accept a context parameter instead of using context.Background(), so Execute's ctx propagates cancellation and timeouts into every volume server RPC - Add deferred cleanup that restores the source volume to writable if any step after markVolumeReadonly fails, preventing the source from being left permanently readonly on abort - Add markVolumeWritable helper using VolumeMarkWritableRequest * fix: deep-copy protobuf messages in test recording sender Use proto.Clone in recordingExecutionSender to store immutable snapshots of JobProgressUpdate and JobCompleted, preventing assertions from observing mutations if the handler reuses message pointers. * fix: add VolumeMarkWritable and ReadVolumeFileStatus to fake volume server The balance task now calls ReadVolumeFileStatus for pre-delete verification and VolumeMarkWritable to restore writability on abort. Add both RPCs to the test fake, and drop the mountCalls assertion since BalanceTask no longer calls VolumeMount directly (VolumeCopy handles it). * fix: use maxConcurrentMovesLimit (50) for clamp, not defaultMaxConcurrentMoves defaultMaxConcurrentMoves (5) is the fallback when the field is unset, not an upper bound. Clamping to it silently overrides valid config values like 10/20/50. Introduce maxConcurrentMovesLimit (50) matching the descriptor's MaxValue and clamp to that instead. * fix: cancel batch moves on progress stream failure Derive a cancellable batchCtx from the caller's ctx. If sender.SendProgress returns an error (client disconnect, context cancelled), capture it, skip further sends, and cancel batchCtx so in-flight moves abort via their propagated context rather than running blind to completion. * fix: bound cleanup timeout and validate batch move fields - Use a 30-second timeout for the deferred markVolumeWritable cleanup instead of context.Background() which can block indefinitely if the volume server is unreachable - Validate required fields (VolumeID, SourceNode, TargetNode) before appending moves to a batch proposal, skipping invalid entries - Fall back to a single-move proposal when filtering leaves only one valid move in a batch * fix: cancel task execution on SendProgress stream failure All handler progress callbacks previously ignored SendProgress errors, allowing tasks to continue executing after the client disconnected. Now each handler creates a derived cancellable context and cancels it on the first SendProgress error, stopping the in-flight task promptly. Handlers fixed: erasure_coding, vacuum, volume_balance (single-move), and admin_script (breaks command loop on send failure). * fix: validate batch moves before scheduling in executeBatchMoves Reject empty batches, enforce a hard upper bound (100 moves), and filter out nil or incomplete move specs (missing source/target/volume) before allocating progress tracking and launching goroutines. * test: add batch balance execution integration test Tests the batch move path with 3 volumes, max concurrency 2, using fake volume servers. Verifies all moves complete with correct readonly, copy, tail, and delete RPC counts. * test: add MarkWritableCount and ReadFileStatusCount accessors Expose the markWritableCalls and readFileStatusCalls counters on the fake volume server, following the existing MarkReadonlyCount pattern. * fix: oscillation guard uses global effective counts for heterogeneous capacity The oscillation guard (max-min <= 1) previously used maxServer/minServer which are determined by utilization ratio. With heterogeneous capacity, maxServer by utilization can have fewer raw volumes than minServer, producing a negative diff and incorrectly triggering the guard. Now scans all servers' effective counts to find the true global max/min volume counts, so the guard works correctly regardless of whether utilization-based or raw-count balancing is used. * fix: admin script handler breaks outer loop on SendProgress failure The break on SendProgress error inside the shell.Commands scan only exited the inner loop, letting the outer command loop continue executing commands on a broken stream. Use a sendBroken flag to propagate the break to the outer execCommands loop.	3 weeks ago
Chris Lu	78a3441b30	fix: volume balance detection returns multiple tasks per run (#8559 ) * fix: volume balance detection now returns multiple tasks per run (#8551) Previously, detectForDiskType() returned at most 1 balance task per disk type, making the MaxJobsPerDetection setting ineffective. The detection loop now iterates within each disk type, planning multiple moves until the imbalance drops below threshold or maxResults is reached. Effective volume counts are adjusted after each planned move so the algorithm correctly re-evaluates which server is overloaded. * fix: factor pending tasks into destination scoring and use UnixNano for task IDs - Use UnixNano instead of Unix for task IDs to avoid collisions when multiple tasks are created within the same second - Adjust calculateBalanceScore to include LoadCount (pending + assigned tasks) in the utilization estimate, so the destination picker avoids stacking multiple planned moves onto the same target disk * test: add comprehensive balance detection tests for complex scenarios Cover multi-server convergence, max-server shifting, destination spreading, pre-existing pending task skipping, no-duplicate-volume invariant, and parameterized convergence verification across different cluster shapes and thresholds. * fix: address PR review findings in balance detection - hasMore flag: compute from len(results) >= maxResults so the scheduler knows more pages may exist, matching vacuum/EC handler pattern - Exhausted server fallthrough: when no eligible volumes remain on the current maxServer (all have pending tasks) or destination planning fails, mark the server as exhausted and continue to the next overloaded server instead of stopping the entire detection loop - Return canonical destination server ID directly from createBalanceTask instead of resolving via findServerIDByAddress, eliminating the fragile address→ID lookup for adjustment tracking - Fix bestScore sentinel: use math.Inf(-1) instead of -1.0 so disks with negative scores (high pending load, same rack/DC) are still selected as the best available destination - Add TestDetection_ExhaustedServerFallsThrough covering the scenario where the top server's volumes are all blocked by pre-existing tasks * test: fix computeEffectiveCounts and add len guard in no-duplicate test - computeEffectiveCounts now takes a servers slice to seed counts for all known servers (including empty ones) and uses an address→ID map from the topology spec instead of scanning metrics, so destination servers with zero initial volumes are tracked correctly - TestDetection_NoDuplicateVolumesAcrossIterations now asserts len > 1 before checking duplicates, so the test actually fails if Detection regresses to returning a single task * fix: remove redundant HasAnyTask check in createBalanceTask The HasAnyTask check in createBalanceTask duplicated the same check already performed in detectForDiskType's volume selection loop. Since detection runs single-threaded (MaxDetectionConcurrency: 1), no race can occur between the two points. * fix: consistent hasMore pattern and remove double-counted LoadCount in scoring - Adopt vacuum_handler's hasMore pattern: over-fetch by 1, check len > maxResults, and truncate — consistent truncation semantics - Remove direct LoadCount penalty in calculateBalanceScore since LoadCount is already factored into effectiveVolumeCount for utilization scoring; bump utilization weight from 40 to 50 to compensate for the removed 10-point load penalty * fix: handle zero maxResults as no-cap, emit trace after trim, seed empty servers - When MaxResults is 0 (omitted), treat as no explicit cap instead of defaulting to 1; only apply the +1 over-fetch probe when caller supplies a positive limit - Move decision trace emission after hasMore/trim so the trace accurately reflects the returned proposals - Seed serverVolumeCounts from ActiveTopology so servers that have a matching disk type but zero volumes are included in the imbalance calculation and MinServerCount check * fix: nil-guard clusterInfo, uncap legacy DetectionFunc, deterministic disk type order - Add early nil guard for clusterInfo in Detection to prevent panics in downstream helpers (detectForDiskType, createBalanceTask) - Change register.go DetectionFunc wrapper from maxResults=1 to 0 (no cap) so the legacy code path returns all detected tasks - Sort disk type keys before iteration so results are deterministic when maxResults spans multiple disk types (HDD/SSD) * fix: don't over-fetch in stateful detection to avoid orphaned pending tasks Detection registers planned moves in ActiveTopology via AddPendingTask, so requesting maxResults+1 would create an extra pending task that gets discarded during trim. Use len(results) >= maxResults as the hasMore signal instead, which is correct since Detection already caps internally. * fix: return explicit truncated flag from Detection instead of approximating Detection now returns (results, truncated, error) where truncated is true only when the loop stopped because it hit maxResults, not when it ran out of work naturally. This eliminates false hasMore signals when detection happens to produce exactly maxResults results by resolving the imbalance. * cleanup: simplify detection logic and remove redundancies - Remove redundant clusterInfo nil check in detectForDiskType since Detection already guards against nil clusterInfo - Remove adjustments loop for destination servers not in serverVolumeCounts — topology seeding ensures all servers with matching disk type are already present - Merge two-loop min/max calculation into a single loop: min across all servers, max only among non-exhausted servers - Replace magic number 100 with len(metrics) for minC initialization in convergence test * fix: accurate truncation flag, deterministic server order, indexed volume lookup - Track balanced flag to distinguish "hit maxResults cap" from "cluster balanced at exactly maxResults" — truncated is only true when there's genuinely more work to do - Sort servers for deterministic iteration and tie-breaking when multiple servers have equal volume counts - Pre-index volumes by server with per-server cursors to avoid O(maxResults * volumes) rescanning on each iteration - Add truncation flag assertions to RespectsMaxResults test: true when capped, false when detection finishes naturally * fix: seed trace server counts from ActiveTopology to match detection logic The decision trace was building serverVolumeCounts only from metrics, missing zero-volume servers seeded from ActiveTopology by Detection. This could cause the trace to report wrong server counts, incorrect imbalance ratios, or spurious "too few servers" messages. Pass activeTopology into the trace function and seed server counts the same way Detection does. * fix: don't exhaust server on per-volume planning failure, sort volumes by ID - When createBalanceTask returns nil, continue to the next volume on the same server instead of marking the entire server as exhausted. The failure may be volume-specific (not found in topology, pending task registration failed) and other volumes on the server may still be viable candidates. - Sort each server's volume slice by VolumeID after pre-indexing so volume selection is fully deterministic regardless of input order. * fix: use require instead of assert to prevent nil dereference panic in CORS test The test used assert.NoError (non-fatal) for GetBucketCors, then immediately accessed getResp.CORSRules. When the API returns an error, getResp is nil causing a panic. Switch to require.NoError/NotNil/Len so the test stops before dereferencing a nil response. * fix: deterministic disk tie-breaking and stronger pre-existing task test - Sort available disks by NodeID then DiskID before scoring so destination selection is deterministic when two disks score equally - Add task count bounds assertion to SkipsPreExistingPendingTasks test: with 15 of 20 volumes already having pending tasks, at most 5 new tasks should be created and at least 1 (imbalance still exists) * fix: seed adjustments from existing pending/assigned tasks to prevent over-scheduling Detection now calls ActiveTopology.GetTaskServerAdjustments() to initialize the adjustments map with source/destination deltas from existing pending and assigned balance tasks. This ensures effectiveCounts reflects in-flight moves, preventing the algorithm from planning additional moves in the same direction when prior moves already address the imbalance. Added GetTaskServerAdjustments(taskType) to ActiveTopology which iterates pending and assigned tasks, decrementing source servers and incrementing destination servers for the given task type.	4 weeks ago
Chris Lu	453310b057	Add plugin worker integration tests for erasure coding (#8450 ) * test: add plugin worker integration harness * test: add erasure coding detection integration tests * test: add erasure coding execution integration tests * ci: add plugin worker integration workflow * test: extend fake volume server for vacuum and balance * test: expand erasure coding detection topologies * test: add large erasure coding detection topology * test: add vacuum plugin worker integration tests * test: add volume balance plugin worker integration tests * ci: run plugin worker tests per worker * fixes * erasure coding: stop after placement failures * erasure coding: record hasMore when early stopping * erasure coding: relax large topology expectations	1 month ago

Author

SHA1

Message

Date

Chris Lu

5f85bf5e8a

Batch volume balance: run multiple moves per job (#8561 )

* proto: add BalanceMoveSpec and batch fields to BalanceTaskParams

Add BalanceMoveSpec message for encoding individual volume moves,
and max_concurrent_moves + repeated moves fields to BalanceTaskParams
to support batching multiple volume moves in a single job.

* balance handler: add batch execution with concurrent volume moves

Refactor Execute() into executeSingleMove() (backward compatible) and
executeBatchMoves() which runs multiple volume moves concurrently using
a semaphore-bounded goroutine pool. When BalanceTaskParams.Moves is
populated, the batch path is taken; otherwise the single-move path.

Includes aggregate progress reporting across concurrent moves,
per-move error collection, and partial failure support.

* balance handler: add batch config fields to Descriptor and worker config

Add max_concurrent_moves and batch_size fields to the worker config
form and deriveBalanceWorkerConfig(). These control how many volume
moves run concurrently within a batch job and the maximum batch size.

* balance handler: group detection proposals into batch jobs

When batch_size > 1, the Detect method groups detection results into
batch proposals where each proposal encodes multiple BalanceMoveSpec
entries in BalanceTaskParams.Moves. Single-result batches fall back
to the existing single-move proposal format for backward compatibility.

* admin UI: add volume balance execution plan and batch badge

Add renderBalanceExecutionPlan() for rich rendering of volume balance
jobs in the job detail modal. Single-move jobs show source/target/volume
info; batch jobs show a moves table with all volume moves.

Add batch badge (e.g., "5 moves") next to job type in the execution
jobs table when the job has batch=true label.

* Update plugin_templ.go

* fix: detection algorithm uses greedy target instead of divergent topology scores

The detection loop tracked effective volume counts via an adjustments map,
but createBalanceTask independently called planBalanceDestination which used
the topology's LoadCount — a separate, unadjusted source of truth. This
divergence caused multiple moves to pile onto the same server.

Changes:
- Add resolveBalanceDestination to resolve the detection loop's greedy
  target (minServer) rather than independently picking a destination
- Add oscillation guard: stop when max-min <= 1 since no single move
  can improve the balance beyond that point
- Track unseeded destinations: if a target server wasn't in the initial
  serverVolumeCounts, add it so subsequent iterations include it
- Add TestDetection_UnseededDestinationDoesNotOverload

* fix: handler force_move propagation, partial failure, deterministic dedupe

- Propagate ForceMove from outer BalanceTaskParams to individual move
  TaskParams so batch moves respect the force_move flag
- Fix partial failure: mark job successful if at least one move
  succeeded (succeeded > 0 || failed == 0) to avoid re-running
  already-completed moves on retry
- Use SHA-256 hash for deterministic dedupe key fallback instead of
  time.Now().UnixNano() which is non-deterministic
- Remove unused successDetails variable
- Extract maxProposalStringLength constant to replace magic number 200

* admin UI: use template literals in balance execution plan rendering

* fix: integration test handles batch proposals from batched detection

With batch_size=20, all moves are grouped into a single proposal
containing BalanceParams.Moves instead of top-level Sources/Targets.
Update assertions to handle both batch and single-move proposal formats.

* fix: verify volume size on target before deleting source during balance

Add a pre-delete safety check that reads the volume file status on both
source and target, then compares .dat file size and file count. If they
don't match, the move is aborted — leaving the source intact rather than
risking irreversible data loss.

Also removes the redundant mountVolume call since VolumeCopy already
mounts the volume on the target server.

* fix: clamp maxConcurrent, serialize progress sends, validate config as int64

- Clamp maxConcurrentMoves to defaultMaxConcurrentMoves before creating
  the semaphore so a stale or malicious job cannot request unbounded
  concurrent volume moves
- Extend progressMu to cover sender.SendProgress calls since the
  underlying gRPC stream is not safe for concurrent writes
- Perform bounds checks on max_concurrent_moves and batch_size in int64
  space before casting to int, avoiding potential overflow on 32-bit

* fix: check disk capacity in resolveBalanceDestination

Skip disks where VolumeCount >= MaxVolumeCount so the detection loop
does not propose moves to a full disk that would fail at execution time.

* test: rename unseeded destination test to match actual behavior

The test exercises a server with 0 volumes that IS seeded from topology
(matching disk type), not an unseeded destination. Rename to
TestDetection_ZeroVolumeServerIncludedInBalance and fix comments.

* test: tighten integration test to assert exactly one batch proposal

With default batch_size=20, all moves should be grouped into a single
batch proposal. Assert len(proposals)==1 and require BalanceParams with
Moves, removing the legacy single-move else branch.

* fix: propagate ctx to RPCs and restore source writability on abort

- All helper methods (markVolumeReadonly, copyVolume, tailVolume,
  readVolumeFileStatus, deleteVolume) now accept a context parameter
  instead of using context.Background(), so Execute's ctx propagates
  cancellation and timeouts into every volume server RPC
- Add deferred cleanup that restores the source volume to writable if
  any step after markVolumeReadonly fails, preventing the source from
  being left permanently readonly on abort
- Add markVolumeWritable helper using VolumeMarkWritableRequest

* fix: deep-copy protobuf messages in test recording sender

Use proto.Clone in recordingExecutionSender to store immutable snapshots
of JobProgressUpdate and JobCompleted, preventing assertions from
observing mutations if the handler reuses message pointers.

* fix: add VolumeMarkWritable and ReadVolumeFileStatus to fake volume server

The balance task now calls ReadVolumeFileStatus for pre-delete
verification and VolumeMarkWritable to restore writability on abort.
Add both RPCs to the test fake, and drop the mountCalls assertion since
BalanceTask no longer calls VolumeMount directly (VolumeCopy handles it).

* fix: use maxConcurrentMovesLimit (50) for clamp, not defaultMaxConcurrentMoves

defaultMaxConcurrentMoves (5) is the fallback when the field is unset,
not an upper bound. Clamping to it silently overrides valid config
values like 10/20/50. Introduce maxConcurrentMovesLimit (50) matching
the descriptor's MaxValue and clamp to that instead.

* fix: cancel batch moves on progress stream failure

Derive a cancellable batchCtx from the caller's ctx. If
sender.SendProgress returns an error (client disconnect, context
cancelled), capture it, skip further sends, and cancel batchCtx so
in-flight moves abort via their propagated context rather than running
blind to completion.

* fix: bound cleanup timeout and validate batch move fields

- Use a 30-second timeout for the deferred markVolumeWritable cleanup
  instead of context.Background() which can block indefinitely if the
  volume server is unreachable
- Validate required fields (VolumeID, SourceNode, TargetNode) before
  appending moves to a batch proposal, skipping invalid entries
- Fall back to a single-move proposal when filtering leaves only one
  valid move in a batch

* fix: cancel task execution on SendProgress stream failure

All handler progress callbacks previously ignored SendProgress errors,
allowing tasks to continue executing after the client disconnected.
Now each handler creates a derived cancellable context and cancels it
on the first SendProgress error, stopping the in-flight task promptly.

Handlers fixed: erasure_coding, vacuum, volume_balance (single-move),
and admin_script (breaks command loop on send failure).

* fix: validate batch moves before scheduling in executeBatchMoves

Reject empty batches, enforce a hard upper bound (100 moves), and
filter out nil or incomplete move specs (missing source/target/volume)
before allocating progress tracking and launching goroutines.

* test: add batch balance execution integration test

Tests the batch move path with 3 volumes, max concurrency 2, using
fake volume servers. Verifies all moves complete with correct readonly,
copy, tail, and delete RPC counts.

* test: add MarkWritableCount and ReadFileStatusCount accessors

Expose the markWritableCalls and readFileStatusCalls counters on the
fake volume server, following the existing MarkReadonlyCount pattern.

* fix: oscillation guard uses global effective counts for heterogeneous capacity

The oscillation guard (max-min <= 1) previously used maxServer/minServer
which are determined by utilization ratio. With heterogeneous capacity,
maxServer by utilization can have fewer raw volumes than minServer,
producing a negative diff and incorrectly triggering the guard.

Now scans all servers' effective counts to find the true global max/min
volume counts, so the guard works correctly regardless of whether
utilization-based or raw-count balancing is used.

* fix: admin script handler breaks outer loop on SendProgress failure

The break on SendProgress error inside the shell.Commands scan only
exited the inner loop, letting the outer command loop continue
executing commands on a broken stream. Use a sendBroken flag to
propagate the break to the outer execCommands loop.

3 weeks ago

Chris Lu

78a3441b30

fix: volume balance detection returns multiple tasks per run (#8559 )

* fix: volume balance detection now returns multiple tasks per run (#8551)

Previously, detectForDiskType() returned at most 1 balance task per disk
type, making the MaxJobsPerDetection setting ineffective. The detection
loop now iterates within each disk type, planning multiple moves until
the imbalance drops below threshold or maxResults is reached. Effective
volume counts are adjusted after each planned move so the algorithm
correctly re-evaluates which server is overloaded.

* fix: factor pending tasks into destination scoring and use UnixNano for task IDs

- Use UnixNano instead of Unix for task IDs to avoid collisions when
  multiple tasks are created within the same second
- Adjust calculateBalanceScore to include LoadCount (pending + assigned
  tasks) in the utilization estimate, so the destination picker avoids
  stacking multiple planned moves onto the same target disk

* test: add comprehensive balance detection tests for complex scenarios

Cover multi-server convergence, max-server shifting, destination
spreading, pre-existing pending task skipping, no-duplicate-volume
invariant, and parameterized convergence verification across different
cluster shapes and thresholds.

* fix: address PR review findings in balance detection

- hasMore flag: compute from len(results) >= maxResults so the scheduler
  knows more pages may exist, matching vacuum/EC handler pattern
- Exhausted server fallthrough: when no eligible volumes remain on the
  current maxServer (all have pending tasks) or destination planning
  fails, mark the server as exhausted and continue to the next
  overloaded server instead of stopping the entire detection loop
- Return canonical destination server ID directly from createBalanceTask
  instead of resolving via findServerIDByAddress, eliminating the
  fragile address→ID lookup for adjustment tracking
- Fix bestScore sentinel: use math.Inf(-1) instead of -1.0 so disks
  with negative scores (high pending load, same rack/DC) are still
  selected as the best available destination
- Add TestDetection_ExhaustedServerFallsThrough covering the scenario
  where the top server's volumes are all blocked by pre-existing tasks

* test: fix computeEffectiveCounts and add len guard in no-duplicate test

- computeEffectiveCounts now takes a servers slice to seed counts for all
  known servers (including empty ones) and uses an address→ID map from
  the topology spec instead of scanning metrics, so destination servers
  with zero initial volumes are tracked correctly
- TestDetection_NoDuplicateVolumesAcrossIterations now asserts len > 1
  before checking duplicates, so the test actually fails if Detection
  regresses to returning a single task

* fix: remove redundant HasAnyTask check in createBalanceTask

The HasAnyTask check in createBalanceTask duplicated the same check
already performed in detectForDiskType's volume selection loop.
Since detection runs single-threaded (MaxDetectionConcurrency: 1),
no race can occur between the two points.

* fix: consistent hasMore pattern and remove double-counted LoadCount in scoring

- Adopt vacuum_handler's hasMore pattern: over-fetch by 1, check
  len > maxResults, and truncate — consistent truncation semantics
- Remove direct LoadCount penalty in calculateBalanceScore since
  LoadCount is already factored into effectiveVolumeCount for
  utilization scoring; bump utilization weight from 40 to 50 to
  compensate for the removed 10-point load penalty

* fix: handle zero maxResults as no-cap, emit trace after trim, seed empty servers

- When MaxResults is 0 (omitted), treat as no explicit cap instead of
  defaulting to 1; only apply the +1 over-fetch probe when caller
  supplies a positive limit
- Move decision trace emission after hasMore/trim so the trace
  accurately reflects the returned proposals
- Seed serverVolumeCounts from ActiveTopology so servers that have a
  matching disk type but zero volumes are included in the imbalance
  calculation and MinServerCount check

* fix: nil-guard clusterInfo, uncap legacy DetectionFunc, deterministic disk type order

- Add early nil guard for clusterInfo in Detection to prevent panics
  in downstream helpers (detectForDiskType, createBalanceTask)
- Change register.go DetectionFunc wrapper from maxResults=1 to 0
  (no cap) so the legacy code path returns all detected tasks
- Sort disk type keys before iteration so results are deterministic
  when maxResults spans multiple disk types (HDD/SSD)

* fix: don't over-fetch in stateful detection to avoid orphaned pending tasks

Detection registers planned moves in ActiveTopology via AddPendingTask,
so requesting maxResults+1 would create an extra pending task that gets
discarded during trim. Use len(results) >= maxResults as the hasMore
signal instead, which is correct since Detection already caps internally.

* fix: return explicit truncated flag from Detection instead of approximating

Detection now returns (results, truncated, error) where truncated is true
only when the loop stopped because it hit maxResults, not when it ran out
of work naturally. This eliminates false hasMore signals when detection
happens to produce exactly maxResults results by resolving the imbalance.

* cleanup: simplify detection logic and remove redundancies

- Remove redundant clusterInfo nil check in detectForDiskType since
  Detection already guards against nil clusterInfo
- Remove adjustments loop for destination servers not in
  serverVolumeCounts — topology seeding ensures all servers with
  matching disk type are already present
- Merge two-loop min/max calculation into a single loop: min across
  all servers, max only among non-exhausted servers
- Replace magic number 100 with len(metrics) for minC initialization
  in convergence test

* fix: accurate truncation flag, deterministic server order, indexed volume lookup

- Track balanced flag to distinguish "hit maxResults cap" from "cluster
  balanced at exactly maxResults" — truncated is only true when there's
  genuinely more work to do
- Sort servers for deterministic iteration and tie-breaking when
  multiple servers have equal volume counts
- Pre-index volumes by server with per-server cursors to avoid
  O(maxResults * volumes) rescanning on each iteration
- Add truncation flag assertions to RespectsMaxResults test: true when
  capped, false when detection finishes naturally

* fix: seed trace server counts from ActiveTopology to match detection logic

The decision trace was building serverVolumeCounts only from metrics,
missing zero-volume servers seeded from ActiveTopology by Detection.
This could cause the trace to report wrong server counts, incorrect
imbalance ratios, or spurious "too few servers" messages. Pass
activeTopology into the trace function and seed server counts the
same way Detection does.

* fix: don't exhaust server on per-volume planning failure, sort volumes by ID

- When createBalanceTask returns nil, continue to the next volume on
  the same server instead of marking the entire server as exhausted.
  The failure may be volume-specific (not found in topology, pending
  task registration failed) and other volumes on the server may still
  be viable candidates.
- Sort each server's volume slice by VolumeID after pre-indexing so
  volume selection is fully deterministic regardless of input order.

* fix: use require instead of assert to prevent nil dereference panic in CORS test

The test used assert.NoError (non-fatal) for GetBucketCors, then
immediately accessed getResp.CORSRules. When the API returns an error,
getResp is nil causing a panic. Switch to require.NoError/NotNil/Len
so the test stops before dereferencing a nil response.

* fix: deterministic disk tie-breaking and stronger pre-existing task test

- Sort available disks by NodeID then DiskID before scoring so
  destination selection is deterministic when two disks score equally
- Add task count bounds assertion to SkipsPreExistingPendingTasks test:
  with 15 of 20 volumes already having pending tasks, at most 5 new
  tasks should be created and at least 1 (imbalance still exists)

* fix: seed adjustments from existing pending/assigned tasks to prevent over-scheduling

Detection now calls ActiveTopology.GetTaskServerAdjustments() to
initialize the adjustments map with source/destination deltas from
existing pending and assigned balance tasks. This ensures
effectiveCounts reflects in-flight moves, preventing the algorithm
from planning additional moves in the same direction when prior
moves already address the imbalance.

Added GetTaskServerAdjustments(taskType) to ActiveTopology which
iterates pending and assigned tasks, decrementing source servers
and incrementing destination servers for the given task type.

4 weeks ago

Chris Lu

453310b057

Add plugin worker integration tests for erasure coding (#8450 )

* test: add plugin worker integration harness

* test: add erasure coding detection integration tests

* test: add erasure coding execution integration tests

* ci: add plugin worker integration workflow

* test: extend fake volume server for vacuum and balance

* test: expand erasure coding detection topologies

* test: add large erasure coding detection topology

* test: add vacuum plugin worker integration tests

* test: add volume balance plugin worker integration tests

* ci: run plugin worker tests per worker

* fixes

* erasure coding: stop after placement failures

* erasure coding: record hasMore when early stopping

* erasure coding: relax large topology expectations

1 month ago

3 Commits (4fe3df6b1d43e80ca5ce249ae6c65b7ecf0ea365)