mirror of
https://github.com/seaweedfs/seaweedfs.git
synced 2026-08-18 05:06:58 +00:00
* fix(topology): drop per-disk task-type conflict map (#9147) Different job types (Balance, ErasureCoding, Vacuum) operate on different volumes, so a per-disk cross-type exclusion adds no correctness guarantee beyond what HasAnyTask already enforces at task detection time. The conflict map turned this into a deadlock on small clusters: a single in-flight (or retrying) balance task would prune the source/destination disks from EC placement, dropping the candidate count below MinTotalDisks and permanently blocking auto-EC. Removing the map lets EC see all eligible disks. Per-volume safety is still guaranteed by HasAnyTask, and per-disk load shaping remains available via MaxConcurrentTasksPerDisk. * chore(topology): trim verbose comments from #9147 fix
88 lines
2.4 KiB
Go
88 lines
2.4 KiB
Go
package topology
|
|
|
|
import (
|
|
"fmt"
|
|
"time"
|
|
)
|
|
|
|
// reassignTaskStates assigns tasks to the appropriate disks
|
|
func (at *ActiveTopology) reassignTaskStates() {
|
|
// Clear existing task assignments
|
|
for _, disk := range at.disks {
|
|
disk.pendingTasks = nil
|
|
disk.assignedTasks = nil
|
|
disk.recentTasks = nil
|
|
}
|
|
|
|
// Reassign pending tasks
|
|
for _, task := range at.pendingTasks {
|
|
at.assignTaskToDisk(task)
|
|
}
|
|
|
|
// Reassign assigned tasks
|
|
for _, task := range at.assignedTasks {
|
|
at.assignTaskToDisk(task)
|
|
}
|
|
|
|
// Reassign recent tasks
|
|
for _, task := range at.recentTasks {
|
|
at.assignTaskToDisk(task)
|
|
}
|
|
}
|
|
|
|
// assignTaskToDisk assigns a task to the appropriate disk(s)
|
|
func (at *ActiveTopology) assignTaskToDisk(task *taskState) {
|
|
addedDisks := make(map[string]bool)
|
|
|
|
// Local helper function to assign task to a disk and avoid code duplication
|
|
assign := func(server string, diskID uint32) {
|
|
key := fmt.Sprintf("%s:%d", server, diskID)
|
|
if server == "" || addedDisks[key] {
|
|
return
|
|
}
|
|
if disk, exists := at.disks[key]; exists {
|
|
switch task.Status {
|
|
case TaskStatusPending:
|
|
disk.pendingTasks = append(disk.pendingTasks, task)
|
|
case TaskStatusInProgress:
|
|
disk.assignedTasks = append(disk.assignedTasks, task)
|
|
case TaskStatusCompleted:
|
|
disk.recentTasks = append(disk.recentTasks, task)
|
|
}
|
|
addedDisks[key] = true
|
|
}
|
|
}
|
|
|
|
// Assign to all source disks
|
|
for _, source := range task.Sources {
|
|
assign(source.SourceServer, source.SourceDisk)
|
|
}
|
|
|
|
// Assign to all destination disks (duplicates automatically avoided by helper)
|
|
for _, dest := range task.Destinations {
|
|
assign(dest.TargetServer, dest.TargetDisk)
|
|
}
|
|
}
|
|
|
|
// isDiskAvailable checks if a disk can accept new tasks. Per-volume safety is
|
|
// enforced by HasAnyTask at detection time, so cross-type tasks on the same
|
|
// disk are intentionally not considered conflicting (see #9147).
|
|
func (at *ActiveTopology) isDiskAvailable(disk *activeDisk, taskType TaskType) bool {
|
|
activeLoad := len(disk.pendingTasks) + len(disk.assignedTasks)
|
|
if MaxConcurrentTasksPerDisk > 0 && activeLoad >= MaxConcurrentTasksPerDisk {
|
|
return false
|
|
}
|
|
return true
|
|
}
|
|
|
|
// cleanupRecentTasks removes old recent tasks
|
|
func (at *ActiveTopology) cleanupRecentTasks() {
|
|
cutoff := time.Now().Add(-time.Duration(at.recentTaskWindowSeconds) * time.Second)
|
|
|
|
for taskID, task := range at.recentTasks {
|
|
if task.CompletedAt.Before(cutoff) {
|
|
delete(at.recentTasks, taskID)
|
|
}
|
|
}
|
|
}
|