Skip to content

fix: skip unhealthy devices in Fit() for all non-nvidia backends #2260

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 8 commits into from
Aug 10, 2026
Merged

fix: skip unhealthy devices in Fit() for all non-nvidia backends #2260

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
4ca2a4e
fix: skip unhealthy devices in Fit() for all non-nvidia backends
Eshiv-Pandey Aug 2, 2026
40dfe0b
fix: address CodeRabbit suggestions and CI lint failures
Eshiv-Pandey Aug 2, 2026
0df0b9d
test: fix broken tests after health check added to Fit()
Eshiv-Pandey Aug 2, 2026
80d4167
test: add Health: true to graphSelect test fixtures after health chec…
Eshiv-Pandey Aug 3, 2026
beb7f68
test: add CardNotHealth coverage for enflame and restore health check…
Eshiv-Pandey Aug 3, 2026
8f67002
refactor(awsneuron): use helper functions for test device construction
Eshiv-Pandey Aug 4, 2026
9b1b530
Merge origin/master into fix/health-check-all-backends
Eshiv-Pandey Aug 4, 2026
3e3fd8a
fix(awsneuron): resolve golangci-lint gofmt and modernize issues
Eshiv-Pandey Aug 4, 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
6 changes: 5 additions & 1 deletion pkg/device/amd/device.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -284,7 +284,11 @@ func (amddevice *AMDDevices) Fit(devices []*device.DeviceUsage, request device.C
for i, v := range slices.Backward(devices) {
dev := v
klog.V(4).InfoS("scoring pod", "pod", klog.KObj(pod), "device", dev.ID, "Memreq", k.Memreq, "MemPercentagereq", k.MemPercentagereq, "Coresreq", k.Coresreq, "Nums", k.Nums, "device index", i)

if !dev.Health {
Comment thread
Shouren marked this conversation as resolved.
reason[common.CardNotHealth]++
klog.V(5).InfoS(common.CardNotHealth, "pod", klog.KObj(pod), "device", dev.ID, "health", dev.Health)
continue
}
klog.V(3).InfoS("Type check", "device", dev.Type, "req", k.Type, "dev=", dev)
_, found, _ := amddevice.checkType(pod.GetAnnotations(), *dev, k)
if !found {
Expand Down
16 changes: 16 additions & 0 deletions pkg/device/amd/device_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -458,4 +458,20 @@ func TestDevices_Fit(t *testing.T) {
assert.Equal(t, false, ok)
assert.Assert(t, strings.Contains(reason, common.CardTimeSlicingExhausted))
})

t.Run("unhealthy device is rejected", func(t *testing.T) {
devices := []*device.DeviceUsage{
{
ID: "dev-0", Index: 0, Used: 0, Count: 2,
Usedmem: 0, Totalmem: 1000, Totalcore: 100, Usedcores: 0,
Type: AMDDevice, Health: false, CustomInfo: map[string]any{},
},
}
req := device.ContainerDeviceRequest{Nums: 1, Type: AMDDevice, Memreq: 100, MemPercentagereq: 0, Coresreq: 10}
pod := &corev1.Pod{ObjectMeta: metav1.ObjectMeta{Annotations: map[string]string{}}}

ok, _, reason := dev.Fit(devices, req, pod, &device.NodeInfo{}, &device.PodDevices{})
assert.Equal(t, false, ok)
assert.Assert(t, strings.Contains(reason, common.CardNotHealth))
})
}
8 changes: 6 additions & 2 deletions pkg/device/awsneuron/device.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -313,7 +313,7 @@ func continuousDeviceAvailable(devices []*device.DeviceUsage, start int, count i
res := []int{}
iterator := start
for iterator < start+count {
if devices[iterator].Used > 0 {
if devices[iterator].Used > 0 || !devices[iterator].Health {
return []int{}
Comment thread
coderabbitai[bot] marked this conversation as resolved.
}
res = append(res, iterator)
Expand Down Expand Up @@ -398,7 +398,11 @@ func (neuron *AWSNeuronDevices) Fit(devices []*device.DeviceUsage, request devic
dev.CustomInfo[AWSUsageInfo] = int(dev.Usedcores)
}
klog.V(4).InfoS("scoring pod", "pod", klog.KObj(pod), "device", dev.ID, "Memreq", k.Memreq, "MemPercentagereq", k.MemPercentagereq, "Coresreq", k.Coresreq, "Nums", k.Nums, "device index", i)

if !dev.Health {
reason[common.CardNotHealth]++
klog.V(5).InfoS(common.CardNotHealth, "pod", klog.KObj(pod), "device", dev.ID, "health", dev.Health)
continue
}
klog.V(3).InfoS("Type check", "device", dev.Type, "req", k.Type, "dev=", dev)
if !strings.Contains(dev.Type, k.Type) {
reason[common.CardTypeMismatch]++
Expand Down
Loading
Loading