Admin: refactoring active topology (#7073)
* refactoring * add ec shard size * address comments * passing task id There seems to be a disconnect between the pending tasks created in ActiveTopology and the TaskDetectionResult returned by this function. A taskID is generated locally and used to create pending tasks via AddPendingECShardTask, but this taskID is not stored in the TaskDetectionResult or passed along in any way. This makes it impossible for the worker that eventually executes the task to know which pending task in ActiveTopology it corresponds to. Without the correct taskID, the worker cannot call AssignTask or CompleteTask on the master, breaking the entire task lifecycle and capacity management feature. A potential solution is to add a TaskID field to TaskDetectionResult and worker_pb.TaskParams, ensuring the ID is propagated from detection to execution. * 1 source multiple destinations * task supports multi source and destination * ec needs to clean up previous shards * use erasure coding constants * getPlanningCapacityUnsafe getEffectiveAvailableCapacityUnsafe should return StorageSlotChange for calculation * use CanAccommodate to calculate * remove dead code * address comments * fix Mutex Copying in Protobuf Structs * use constants * fix estimatedSize The calculation for estimatedSize only considers source.EstimatedSize and dest.StorageChange, but omits dest.EstimatedSize. The TaskDestination struct has an EstimatedSize field, which seems to be ignored here. This could lead to an incorrect estimation of the total size of data involved in tasks on a disk. The loop should probably also include estimatedSize += dest.EstimatedSize. * at.assignTaskToDisk(task) * refactoring * Update weed/admin/topology/internal.go Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> * fail fast * fix compilation * Update weed/worker/tasks/erasure_coding/detection.go Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> * indexes for volume and shard locations * dedup with ToVolumeSlots * return an additional boolean to indicate success, or an error * Update abstract_sql_store.go * fix * Update weed/worker/tasks/erasure_coding/detection.go Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> * Update weed/admin/topology/task_management.go Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> * faster findVolumeDisk * Update weed/worker/tasks/erasure_coding/detection.go Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * Update weed/admin/topology/storage_slot_test.go Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * refactor * simplify * remove unused GetDiskStorageImpact function * refactor * add comments * Update weed/admin/topology/storage_impact.go Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> * Update weed/admin/topology/storage_slot_test.go Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> * Update storage_impact.go * AddPendingTask The unified AddPendingTask function now serves as the single entry point for all task creation, successfully consolidating the previously separate functions while maintaining full functionality and improving code organization. --------- Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
This commit is contained in:
Chris Lu
@@ -83,7 +83,11 @@ func Detection(metrics []*types.VolumeHealthMetrics, clusterInfo *types.ClusterI
|
||||
reason := fmt.Sprintf("Cluster imbalance detected: %.1f%% (max: %d on %s, min: %d on %s, avg: %.1f)",
|
||||
imbalanceRatio*100, maxVolumes, maxServer, minVolumes, minServer, avgVolumesPerServer)
|
||||
|
||||
// Generate task ID for ActiveTopology integration
|
||||
taskID := fmt.Sprintf("balance_vol_%d_%d", selectedVolume.VolumeID, time.Now().Unix())
|
||||
|
||||
task := &types.TaskDetectionResult{
|
||||
TaskID: taskID, // Link to ActiveTopology pending task
|
||||
TaskType: types.TaskTypeBalance,
|
||||
VolumeID: selectedVolume.VolumeID,
|
||||
Server: selectedVolume.Server,
|
||||
@@ -103,6 +107,7 @@ func Detection(metrics []*types.VolumeHealthMetrics, clusterInfo *types.ClusterI
|
||||
|
||||
// Create typed parameters with destination information
|
||||
task.TypedParams = &worker_pb.TaskParams{
|
||||
TaskId: taskID, // Link to ActiveTopology pending task
|
||||
VolumeId: selectedVolume.VolumeID,
|
||||
Server: selectedVolume.Server,
|
||||
Collection: selectedVolume.Collection,
|
||||
@@ -121,6 +126,35 @@ func Detection(metrics []*types.VolumeHealthMetrics, clusterInfo *types.ClusterI
|
||||
|
||||
glog.V(1).Infof("Planned balance destination for volume %d: %s -> %s (score: %.2f)",
|
||||
selectedVolume.VolumeID, selectedVolume.Server, destinationPlan.TargetNode, destinationPlan.PlacementScore)
|
||||
|
||||
// Add pending balance task to ActiveTopology for capacity management
|
||||
|
||||
// Find the actual disk containing the volume on the source server
|
||||
sourceDisk, found := base.FindVolumeDisk(clusterInfo.ActiveTopology, selectedVolume.VolumeID, selectedVolume.Collection, selectedVolume.Server)
|
||||
if !found {
|
||||
return nil, fmt.Errorf("BALANCE: Could not find volume %d (collection: %s) on source server %s - unable to create balance task",
|
||||
selectedVolume.VolumeID, selectedVolume.Collection, selectedVolume.Server)
|
||||
}
|
||||
targetDisk := destinationPlan.TargetDisk
|
||||
|
||||
err = clusterInfo.ActiveTopology.AddPendingTask(topology.TaskSpec{
|
||||
TaskID: taskID,
|
||||
TaskType: topology.TaskTypeBalance,
|
||||
VolumeID: selectedVolume.VolumeID,
|
||||
VolumeSize: int64(selectedVolume.Size),
|
||||
Sources: []topology.TaskSourceSpec{
|
||||
{ServerID: selectedVolume.Server, DiskID: sourceDisk},
|
||||
},
|
||||
Destinations: []topology.TaskDestinationSpec{
|
||||
{ServerID: destinationPlan.TargetNode, DiskID: targetDisk},
|
||||
},
|
||||
})
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("BALANCE: Failed to add pending task for volume %d: %v", selectedVolume.VolumeID, err)
|
||||
}
|
||||
|
||||
glog.V(2).Infof("Added pending balance task %s to ActiveTopology for volume %d: %s:%d -> %s:%d",
|
||||
taskID, selectedVolume.VolumeID, selectedVolume.Server, sourceDisk, destinationPlan.TargetNode, targetDisk)
|
||||
} else {
|
||||
glog.Warningf("No ActiveTopology available for destination planning in balance detection")
|
||||
return nil, nil
|
||||
|
||||
Reference in New Issue
Block a user