Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Fix the scheduler panic whenever the GPU is lost on node #306

Merged
merged 5 commits into from
Jul 8, 2019
Merged

Fix the scheduler panic whenever the GPU is lost on node #306

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
a619c2f
Merge pull request #1 from volcano-sh/master
william-wang Jun 19, 2019
03120f0
Merge pull request #2 from volcano-sh/master
william-wang Jun 28, 2019
da7e392
Merge remote-tracking branch 'upstream/master'
william-wang Jul 8, 2019
518da33
Fix scheduler panic happens when the GPU is lost on node
william-wang Jul 8, 2019
52d9732
Fix the build issue
william-wang Jul 8, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 18 additions & 2 deletions pkg/scheduler/api/node_info.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -161,6 +161,18 @@ func (ni *NodeInfo) SetNode(node *v1.Node) {
}
}

func (ni *NodeInfo) allocateIdleResource(ti *TaskInfo) error {
if ti.Resreq.LessEqual(ni.Idle) {
ni.Idle.Sub(ti.Resreq)
return nil
}
ni.State = NodeState{
Phase: NotReady,
Reason: "OutOfSync",
}
return fmt.Errorf("Selected node NotReady")
}

// AddTask is used to add a task in nodeInfo object
func (ni *NodeInfo) AddTask(task *TaskInfo) error {
key := PodKey(task.Pod)
Expand All @@ -176,12 +188,16 @@ func (ni *NodeInfo) AddTask(task *TaskInfo) error {
if ni.Node != nil {
switch ti.Status {
case Releasing:
if err := ni.allocateIdleResource(ti); err != nil {
return err
}
ni.Releasing.Add(ti.Resreq)
ni.Idle.Sub(ti.Resreq)
case Pipelined:
ni.Releasing.Sub(ti.Resreq)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Does this face the same problem?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, So I deleted this line and replace it to line 191-193.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@william-wang Could you add more comments why this can cause a panic?
In customer's environment, when the GPU was lost. the ni.Idle decreased and the value was less than ti.Resreq. Then the panic happened in Sub func. So we need to add a judgement before the execution of ni.Idle.Sub(ti.Resreq).

default:
ni.Idle.Sub(ti.Resreq)
if err := ni.allocateIdleResource(ti); err != nil {
return err
}
}

ni.Used.Add(ti.Resreq)
Expand Down
19 changes: 19 additions & 0 deletions pkg/scheduler/api/node_info_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,9 @@ func TestNodeInfo_AddPod(t *testing.T) {
case01Node := buildNode("n1", buildResourceList("8000m", "10G"))
case01Pod1 := buildPod("c1", "p1", "n1", v1.PodRunning, buildResourceList("1000m", "1G"), []metav1.OwnerReference{}, make(map[string]string))
case01Pod2 := buildPod("c1", "p2", "n1", v1.PodRunning, buildResourceList("2000m", "2G"), []metav1.OwnerReference{}, make(map[string]string))
// case2
case02Node := buildNode("n2", buildResourceList("2000m", "1G"))
case02Pod1 := buildPod("c2", "p1", "n2", v1.PodUnknown, buildResourceList("1000m", "2G"), []metav1.OwnerReference{}, make(map[string]string))

tests := []struct {
name string
Expand All @@ -63,6 +66,22 @@ func TestNodeInfo_AddPod(t *testing.T) {
},
},
},
{
name: "add 1 unknown pod",
node: case02Node,
pods: []*v1.Pod{case02Pod1},
expected: &NodeInfo{
Name: "n2",
Node: case02Node,
Idle: buildResource("2000m", "1G"),
Used: EmptyResource(),
Releasing: EmptyResource(),
Allocatable: buildResource("2000m", "1G"),
Capability: buildResource("2000m", "1G"),
State: NodeState{Phase: NotReady, Reason: "OutOfSync"},
Tasks: map[TaskID]*TaskInfo{},
},
},
}

for i, test := range tests {
Expand Down