Skip to content

fix(ascend): stop whole-card/memory-less hami-core requests from bypassing the node mode gate #2029

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

Merged
hami-robot merged 3 commits into from
Jul 9, 2026
+82 −6
Merged

fix(ascend): stop whole-card/memory-less hami-core requests from bypassing the node mode gate #2029

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
f9830b0
fix(ascend): stop whole-card/memory-less hami-core requests from bypa…
Jul 7, 2026
6a4a46f
address review: make the legacy-vNPU filter log nil-safe
Jul 8, 2026
cdbaaae
address review: drop verbose comments and internal markers
Jul 8, 2026
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
14 changes: 8 additions & 6 deletions pkg/device/ascend/device.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -460,14 +460,16 @@ func (npu *Devices) Fit(devices []*device.DeviceUsage, request device.ContainerD
totalMemPerCard = devices[0].Totalmem
}

if isHAMiCore && !nodeSupportHamiCore {
reason[common.ModeNotFit]++
klog.V(4).InfoS("Node filtered: pod requests hami-core but node does not support it", "pod", klog.KObj(pod))
return false, nil, common.GenReason(reason, len(devices))
}

if request.Memreq > 0 && request.Memreq < totalMemPerCard && request.Nums > 0 {
if !nodeSupportHamiCore && isHAMiCore {
reason[common.ModeNotFit]++
klog.V(4).InfoS("Node filtered: Node does not support hami-core mode", "node", nodeInfo.Node.Name, "pod", pod.Name)
return false, nil, common.GenReason(reason, len(devices))
} else if nodeSupportHamiCore && !isHAMiCore {
if nodeSupportHamiCore && !isHAMiCore {
reason[common.ModeNotFit]++
klog.V(4).InfoS("Node filtered: Reserved for hami-core but pod is legacy vNPU", "node", nodeInfo.Node.Name, "pod", pod.Name)
klog.V(4).InfoS("Node filtered: node reserved for hami-core but pod is legacy vNPU", "pod", klog.KObj(pod))
return false, nil, common.GenReason(reason, len(devices))
Comment thread
Wangmin362 marked this conversation as resolved.
}
}
Expand Down
74 changes: 74 additions & 0 deletions pkg/device/ascend/device_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1813,6 +1813,80 @@ func TestDevices_Fit(t *testing.T) {
wantReason: "1/1 ModeNotFit",
nodeAnnotation: map[string]string{VNPUNodeSelectorAnnotation: "true"},
},
{
name: "fit fail: whole-card hami-core pod on legacy node (ModeNotFit)",
devices: []*device.DeviceUsage{{
ID: "dev-0", Index: 0, Used: 0, Count: 100,
Usedmem: 0, Totalmem: 32768, Totalcore: 100, Usedcores: 0,
Numa: 0, Health: true,
}},
request: device.ContainerDeviceRequest{
Nums: 1, Memreq: 32768, MemPercentagereq: 0, Coresreq: 0,
},
annos: map[string]string{
VNPUModeAnnotation: VNPUModeHamiCore,
},
wantFit: false,
wantLen: 0,
wantDevIDs: []string{},
wantReason: "1/1 ModeNotFit",
nodeAnnotation: map[string]string{},
},
{
name: "fit fail: memory-less hami-core pod on legacy node (ModeNotFit)",
devices: []*device.DeviceUsage{{
ID: "dev-0", Index: 0, Used: 0, Count: 100,
Usedmem: 0, Totalmem: 32768, Totalcore: 100, Usedcores: 0,
Numa: 0, Health: true,
}},
request: device.ContainerDeviceRequest{
Nums: 1, Memreq: 0, MemPercentagereq: 101, Coresreq: 0,
},
annos: map[string]string{
VNPUModeAnnotation: VNPUModeHamiCore,
},
wantFit: false,
wantLen: 0,
wantDevIDs: []string{},
wantReason: "1/1 ModeNotFit",
nodeAnnotation: map[string]string{},
},
{
name: "fit success: whole-card hami-core pod on hami-core node",
devices: []*device.DeviceUsage{{
ID: "dev-0", Index: 0, Used: 0, Count: 100,
Usedmem: 0, Totalmem: 32768, Totalcore: 100, Usedcores: 0,
Numa: 0, Health: true,
}},
request: device.ContainerDeviceRequest{
Nums: 1, Memreq: 32768, MemPercentagereq: 0, Coresreq: 0,
},
annos: map[string]string{
VNPUModeAnnotation: VNPUModeHamiCore,
},
wantFit: true,
wantLen: 1,
wantDevIDs: []string{"dev-0"},
wantReason: "",
nodeAnnotation: map[string]string{VNPUNodeSelectorAnnotation: "true"},
},
{
name: "fit success: whole-card legacy pod on hami-core node",
devices: []*device.DeviceUsage{{
ID: "dev-0", Index: 0, Used: 0, Count: 100,
Usedmem: 0, Totalmem: 32768, Totalcore: 100, Usedcores: 0,
Numa: 0, Health: true,
}},
request: device.ContainerDeviceRequest{
Nums: 1, Memreq: 32768, MemPercentagereq: 0, Coresreq: 0,
},
annos: map[string]string{},
wantFit: true,
wantLen: 1,
wantDevIDs: []string{"dev-0"},
wantReason: "",
nodeAnnotation: map[string]string{VNPUNodeSelectorAnnotation: "true"},
},
}

for _, dev := range devs {
Expand Down
Loading