synchronization

gpushare-scheduler-extender/deployer/docker/device-plugin-recover/Dockerfile

@@ -0,0 +1,12 @@
+FROM debian:bullseye-slim
+RUN echo \
+    deb [arch=amd64]  http://mirrors.aliyun.com/debian/ bullseye main non-free contrib\
+    > /etc/apt/sources.list
+RUN apt-get update
+RUN apt-get install -y curl tzdata iproute2 bash &&  \
+ 	rm -rf /var/cache/apt/* && \
+ 	cp /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && \
+ 	echo "Asia/Shanghai" >  /etc/timezone && \
+ 	mkdir -p /dp-evict
+ADD dp-evict /dp-evict
+RUN chmod -R +x /dp-evict

gpushare-scheduler-extender/deployer/docker/device-plugin-recover/build.sh

@@ -0,0 +1,16 @@
+#!/usr/bin/env bash
+set -xe
+
+BASEDIR=$(dirname "$0")
+echo "build docker images in $BASEDIR"
+
+TIMESTAMP=$(date +%Y%m%d%H%M)
+
+cd $BASEDIR
+
+# docker build --no-cache -t $IMAGE -f $FILE $BASEDIR
+docker build --network=host -t registry.cn-hangzhou.aliyuncs.com/acs/nvidia-device-plugin-recover:$TIMESTAMP .
+
+docker push registry.cn-hangzhou.aliyuncs.com/acs/nvidia-device-plugin-recover:$TIMESTAMP
+
+echo registry.cn-hangzhou.aliyuncs.com/acs/nvidia-device-plugin-recover:$TIMESTAMP

gpushare-scheduler-extender/deployer/docker/device-plugin-recover/dp-evict/dp-recover-on-host.sh

@@ -0,0 +1,34 @@
+#!/usr/bin/env bash
+
+set -e -x
+
+dir="/etc/kubernetes/manifests"
+backup_dir="/etc/kubernetes/manifests-backup"
+
+public::common::log() {
+	echo $(date +"[%Y%m%d %H:%M:%S]: ") $1
+}
+
+public::recover::gpu-device-plugin() {
+
+    if [  -f $dir/nvidia-device-plugin.yml ]; then
+		public::common::log "Skip recovering nvidia-device-plugin.yml, because it already exist."
+    else
+    	if [  -f $backup_dir/nvidia-device-plugin.yml ]; then
+			cp -f $backup_dir/nvidia-device-plugin.yml $dir/nvidia-device-plugin.yml
+			public::common::log "Finish recovering nvidia-device-plugin.yml."
+		else
+			public::common::log "No nvidia-device-plugin.yml to recover."
+		fi	
+	fi
+}
+
+main() {
+
+	public::recover::gpu-device-plugin
+
+	touch /ready
+	while sleep 3600; do :; done
+}
+
+main "$@"

gpushare-scheduler-extender/deployer/docker/device-plugin-recover/dp-evict/dp-recover.sh

@@ -0,0 +1,11 @@
+#!/usr/bin/env bash
+set -xe
+
+if [ -d "/k8s-host" ]; then
+	rm -rf /k8s-host/usr/local/dp-evict
+	mkdir -p /k8s-host/usr/local/dp-evict
+	cp -r /dp-evict/* /k8s-host/usr/local/dp-evict
+	chmod -R +x /k8s-host/usr/local/dp-evict
+	chroot /k8s-host /usr/local/dp-evict/dp-recover-on-host.sh "$@"
+	while sleep 3600; do :; done
+fi