From dd92a413cef036188f666a7e1f8922e8b2017cae Mon Sep 17 00:00:00 2001
From: ghokun <gokhun@gmail.com>
Date: Mon, 21 Mar 2022 23:15:15 +0300
Subject: [PATCH] This will go on forever

Signed-off-by: ghokun <gokhun@gmail.com>
---
 manifests/device-plugin.yml |  2 +-
 pkg/gpu/nvidia/metrics.go   | 29 +++++++++++++++--------------
 2 files changed, 16 insertions(+), 15 deletions(-)

diff --git a/manifests/device-plugin.yml b/manifests/device-plugin.yml
index b732e50..f6cb4b0 100644
--- a/manifests/device-plugin.yml
+++ b/manifests/device-plugin.yml
@@ -45,7 +45,7 @@ spec:
             capabilities:
               add: ["SYS_ADMIN"]
       containers:
-        - image: ghcr.io/kuartis/kuartis-virtual-gpu-device-plugin:0.4.9
+        - image: ghcr.io/kuartis/kuartis-virtual-gpu-device-plugin:0.4.10
           name: kuartis-virtual-gpu-device-plugin-ctr
           command:
             - /usr/bin/virtual-gpu-device-plugin
diff --git a/pkg/gpu/nvidia/metrics.go b/pkg/gpu/nvidia/metrics.go
index 7d50515..cbb11c7 100644
--- a/pkg/gpu/nvidia/metrics.go
+++ b/pkg/gpu/nvidia/metrics.go
@@ -96,20 +96,21 @@ func collectMetrics(w http.ResponseWriter, r *http.Request) {
 		log.Printf("Found %d processes on GPU %d", len(processes), i)
 		for _, process := range processes {
 			containerId := getContainerId(process.Pid)
-			container := containerMap[containerId]
-			log.Printf("Using %s Found container %+v for process: %d", containerId, container, process.Pid)
-			collected = append(collected, metric{
-				Pid:           process.Pid,
-				UsedGpuMemory: process.UsedGpuMemory,
-				GpuIndex:      i,
-				GpuUUID:       getDeviceUUID(d),
-				Node:          container.Node,
-				Namespace:     container.Namespace,
-				Pod:           container.Pod,
-				PodUid:        container.PodUid,
-				Container:     container.Container,
-				ContainerId:   container.ContainerId,
-			})
+			if container, ok := containerMap[strings.TrimSpace(containerId)]; ok {
+				log.Printf("Using %s Found container %+v for process: %d", containerId, container, process.Pid)
+				collected = append(collected, metric{
+					Pid:           process.Pid,
+					UsedGpuMemory: process.UsedGpuMemory,
+					GpuIndex:      i,
+					GpuUUID:       getDeviceUUID(d),
+					Node:          container.Node,
+					Namespace:     container.Namespace,
+					Pod:           container.Pod,
+					PodUid:        container.PodUid,
+					Container:     container.Container,
+					ContainerId:   container.ContainerId,
+				})
+			}
 		}
 	}