YAO/YAO-scheduler
1
0
Fork 0
mirror of https://github.com/newnius/YAO-scheduler.git synced 2025-06-07 14:21:55 +00:00
Code Issues Releases Wiki Activity

robust, make sure allocatinngGPU is decreased even if job is killed before requesting resource

Browse Source
This commit is contained in:
Newnius 2020-07-11 11:17:38 +08:00
parent ee34e09f7f
commit 71ab7bbb75
1 changed files with 5 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
src/scheduler_fair.go
View File

@ -367,7 +367,11 @@ func (scheduler *SchedulerFair) UpdateProgress(job Job, state State) {
defer scheduler.historyMu.Unlock() defer scheduler.historyMu.Unlock()
scheduler.schedulingMu.Lock() scheduler.schedulingMu.Lock()
delete(scheduler.schedulingJobs, job.Name) if _, ok := scheduler.schedulingJobs[job.Name]; ok {
delete(scheduler.schedulingJobs, job.Name)
scheduler.allocatingGPU -= job.NumberGPU
log.Info("allocatingGPU is ", scheduler.allocatingGPU)
}
scheduler.schedulingMu.Unlock() scheduler.schedulingMu.Unlock()
switch state { switch state {
@ -456,15 +460,6 @@ func (scheduler *SchedulerFair) Schedule(job Job) {
func (scheduler *SchedulerFair) AcquireResource(job Job) []NodeStatus { func (scheduler *SchedulerFair) AcquireResource(job Job) []NodeStatus {
res := InstanceOfResourcePool().acquireResource(job) res := InstanceOfResourcePool().acquireResource(job)
if len(res) != 0 {
for _, task := range job.Tasks {
scheduler.allocatingGPUMu.Lock()
scheduler.allocatingGPU -= task.NumberGPU
scheduler.allocatingGPUMu.Unlock()
}
log.Info("allocatingGPU is ", scheduler.allocatingGPU)
}
go func() { go func() {
scheduler.UpdateQuota() scheduler.UpdateQuota()
}() }()