vah101的专栏

head-less请求的例子如下：apiVersion: apps/v1kind: StatefulSetmetadata: labels: app: swm name: swmspec: replicas: 2 selector: matchLabels: app: swm serviceName: swm-headless template: metadata: labels: app: swm

centos8版本(linux内核版本高于3.13)采用了nftables而不是iptables做数据包的转发，所以使用默认的iptables方式可能会出现错误。需要修改集群配置，修改集群配置文件为如下形势： kubeproxy: extra_args: # 默认使用iptables进行数据转发 proxy-mode: "ipvs" # 如果要启用ipvs，则此处设置为`ipvs`

原因如下：A problem in BIRD-Linux kernel routing table synchronization when BIRD tries to overwrite an existing kernel route. There are two common causes:First, there are some routes in the kernel routing table added by some other tools (like ip or route co

virsh list 显示正在运行的虚拟机virsh list --all显示所有的虚拟机，包括已经关闭的virsh dumpxml xxx > ./vm.xml 将虚拟机描述文件导出virsh define vm.xml 导入虚拟机描述文件virsh start xxx 启动虚拟机xxxvirsh destroy xxx 关闭虚拟机xxx...

cat >> /usr/lib/systemd/system/harbor.service <<EOF[Unit]Description=HarborAfter=docker.service systemd-networkd.service systemd-resolved.serviceRequires=docker.serviceDocumentation=http://github.com/vmware/harbor[Service]Type=simpleR.

apiVersion: extensions/v1beta1kind: Ingressmetadata: annotations: nginx.ingress.kubernetes.io/cors-allow-methods: '*' nginx.ingress.kubernetes.io/cors-allow-origin: '*' nginx.ingress.kubernetes.io/enable-cors: "true" nginx.ingress.kube.

calico日志中间歇性输出如下内容：2021-07-03 10:02:15.879 [INFO][100] iface_monitor.go 176: Netlink address update. addr="192.168.1.150" exists=false ifIndex=92021-07-03 10:02:15.879 [INFO][100] int_dataplane.go 622: Linux interface addrs changed. addrs=set.mapSet{"1

连接ignite数据库./sqlline.sh --verbose=true -u jdbc:ignite:thin://10.42.246.146/创建表：CREATE TABLE PUBLIC.Person (id INTEGER,NAME VARCHAR,PRIMARY KEY (id));生成测试sql脚本for ((i=1; i<=100000; i ++)) do echo "insert into PUBLIC.Person (id, NAME) values($...

编辑ignite.yaml文件如下：apiVersion: v1kind: Servicemetadata: # Name of Ignite Service used by Kubernetes IP finder. # The name must be equal to TcpDiscoveryKubernetesIpFinder.serviceName. name: ignite namespace: defaultspec: clusterIP: None # cu

1. 获取操作系统镜像：从https://cloud.centos.org/centos/7/images/下载qcow2格式的操作系统镜像2. 执行如下操作：kubectl -n cdi get svc -l cdi.kubevirt.io=cdi-uploadproxy获取cdi-uploadproxy的ip地址，如10.43.138.413. 编写pv.yaml内容如下：apiVersion: v1kind: PersistentVolumemetadata: n.

首先要安装jq：wget http://dl.fedoraproject.org/pub/epel/epel-release-latest-7.noarch.rpmrpm -ivh epel-release-latest-7.noarch.rpmyum install jq设置要删除命名空间的名称：NAMESPACE=test接着执行如下操作：RANCHER_SERVER_URL=$( kubectl config view -o json|jq -r .clusters[0

将原有集群删除后，重新创建集群，rancher集群报错：Cluster health check failed: Failed to communicate with API server: Get "https://192.168.200.10:6443/api/v1/namespaces/kube-system?timeout=45s": context deadline exceeded 检查6443对应api-server的docker容器的日志：docker logs -f kub

FROM nvidia/cuda:10.0-cudnn7-devel-centos7# install basic dependenciesRUN yum install vim wget cmake bzip2 -y# install Anaconda3RUN wget https://mirrors.tuna.tsinghua.edu.cn/anaconda/archive/Anaconda3-5.3.1-Linux-x86_64.sh -O ~/Anaconda3-5.3.1-L...

http://greenplum-kubernetes.docs.pivotal.io/2-3/index.html部署监控程序GPCChttps://greenplum-kubernetes.docs.pivotal.io/2-3/gpcc.html

calico提示：Invalidating dataplane cache ipVersion=0x4 reason="chain update" table="filter"解决方法，修改calico对应yml，修改readness和liveness对应的内容为： livenessProbe: failureThreshold: 6 httpGet: host: localhost pat

calico报错： int_dataplane.go 1018: Kernel's RPF check is set to 'loose'. This would allow endpoints to spoof their IP address. Calico requires net.ipv4.conf.all.rp_filter to be set to 0 or 1. If you require loose RPF and you are not concerned about spoofi

对于多个网口的服务器，calico需要指定网口，可以在calico.yaml的containers->env中，加入环境变量IP_AUTODETECTION_METHOD： containers: - env: - name: IP_AUTODETECTION_METHOD value: interface=bond*,eth*其中interface可以使用通配符，并且可以指定多种类型的前缀...

1. 表现是rancher界面无法登陆，通过docker logs -f rancher发现日志中主要是etcd在报错，包含如下信息：etcdserver: mvcc: database space exceeded2. 确定是由于etcd写满限额，导致的问题。需要登录到etcd所在的结点执行：docker exec -ti etcd sh进入到etcd的容器，再执行，注意需要进入到每个etcd结点依次执行：#查看etcd的告警信息etcdctl alarm list#获得版

https://blog.youkuaiyun.com/vah101/article/details/109393495介绍了部署kubevirt实现基于k8s的kvm虚拟化，但是这里提到的kvm虚机并没实现持久化。kubevirt提供了一种持久化方式Containerized Data Importer (CDI)。其构建过程为：wget https://raw.githubusercontent.com/kubevirt/kubevirt.github.io/master/labs/manifes...

1. 安装相关的镜像，执行如下命令export VERSION=$(curl -s https://api.github.com/repos/kubevirt/kubevirt/releases | grep tag_name | grep -v -- '-rc' | head -1 | awk -F': ' '{print $2}' | sed 's/,//' | xargs)echo $VERSIONkubectl create -f https://github.com/kubevirt/k

按https://blog.youkuaiyun.com/vah101/article/details/108828526操作，更换config文件后，再执行virtctl console就正常了

在rancher下创建证书后，使用kubectl describe csr xxx查询状态为pending，之后执行kubectl certificate approve xxx将证书批准，再检查状态，发现变为Approved，但是执行：kubectl get csr xxx -o jsonpath='{.status.certificate}'没有输出结果，再次执行kubectl describe csr xxx仔细观察，发现status只是Approved，正常情况下，应该为Ap

https://blog.youkuaiyun.com/cd_yourheart/article/details/107463956

k8s安装tidb如果是用Rancher管理的集群，需要修改集群配置挂载分区启动local-volume-provisioner创建crd器安装helm安装tidb-operator安装tidb-cluster获取tidb的nodePort连接tidbEnd# tidb-operater要求集群中最少有3个结点如果是用Rancher管理的集群，需要修改集群配置需要到集群->升级->编辑yaml下，找到“ services:”，在其中为kubelet增加参数，如下： kubelet:

在docker中运行的pytorch运行多线程训练，报如下错误：unexpected bus error encountered in worker. This might be caused by insufficient shared memory(shm)解决方法，启动docker时，增加--ipc=host参数如果是在k8s上运行的，则需要在yaml的spec中加入hostIPC: true类似如下效果：apiVersion: v1kind: podmetadata:.

rancher运行需要配置pv、pvc的应用如下报错：Warning FailedMount 2s (x6 over 17s) kubelet, node1 MountVolume.NewMounter initialization failed for volume "local-pv-cadb07da" : path "/mnt/ssd/t1" does not exist需要到集群->升级->编辑yaml下，找到“services:”，在其中为kubelet增加参...

pytorch查看GPU信息import torchtorch.cuda.is_available()#cuda是否可用；torch.cuda.device_count()#返回gpu数量；torch.cuda.get_device_name(0)#返回gpu名字，设备索引默认从0开始；torch.cuda.current_device()tensorflow查看GPU信息：import tensorflow as tffrom tensorflow.python.

清除掉已经停止的docker容器docker rm `docker ps -a|grep 'Exited'|awk {'print $1'}`清除掉已经被代替的docker镜像docker rmi `docker images|grep '<none>'|awk {'print $3'}`

参考：https://support.huaweicloud.com/instg-kunpengcpfs/kunpengcpfs_03_0001.html1. 到这里下载2进制的docker可执行文件https://download.docker.com/linux/static/stable/aarch64/2. 解压缩tar xvpf docker-19.03.8.tgz cp -p docker/* /usr/binsetenforce 0systemctl st.

rancher界面报错：[workerPlane] Failed to bring up Worker Plane: [Failed to verify healthcheck: Failed to check http://localhost:10248/healthz for service [kubelet] on host [192.168.0.101]: Get http://localhost:10248/healthz: dial tcp [::1]:10248: connect: co

使用kubectl creat -f执行一个yaml脚本的过程中报错：[debug] error converting YAML to JSON: yaml: line 42: did not find expected ':' indicator后来仔细检查，错误原因是不同段落之前的分隔符本应该是"---"(三个减号)，少写了一个，只有两个。将分隔符补齐，报错提示消失。...

编写如下的tensflow.yaml文件apiVersion: apps/v1kind: Deploymentmetadata: name: tensorflow-gpu-jupyter labels: app: tensorflow-gpu-jupyterspec: replicas: 1 selector: # define how the deployment finds the pods it mangages matchLabels: app

gpushare-scheduler-extender是阿里云在kubernetes平台上开发的针对GPU进行虚拟化的方案，首先，参考https://blog.youkuaiyun.com/vah101/article/details/108098827，安装k8s-deviece-plugin，并将/etc/docker/daemon.json配置为：{ "default-runtime": "nvidia", "runtimes": { "nvidia": {

首先，下载etcd代码mkdir -p $GOPATH/src/github.com/coreoscd $GOPATH/src/github.com/coreosgit clone https://github.com/etcd-io/etcd.gitcd etcdgit chechout v3.3.15make build如果在arm64环境下，出现如下结果，说明成功2020-08-20 18:05:34.183276 E | etcdmain: etcd on unsuppo

1.在GPU服务器上安装cuda程序及驱动执行:lspci|grep-invidia确定是否存在GPU，如果提示lspci命令不存在，则执行yuminstallpciutils-y2.安装NVIDIA、epel的rpm仓库执行：wgethttps://developer.download.nvidia.cn/compute/cuda/repos/rhel7/x86_64/cuda-repo-rhel7-10.1.243-1.x86_6...

2.2版本的rancher证书过期，docker会频繁重启报错：08:51:46.160121 I | http: TLS handshake error from 127.0.0.1:33140: remote error: tls: bad certificateE0814 08:51:46.160212 6 reflector.go:134] k8s.io/client-go/informers/factory.go:127: Failed to list *v1.Replicat

kubelet容器频繁启动，日志报错：kubelet.go:1344] Failed to start cAdvisor inotify_add_watch /sys/fs/cgroup/devices: no space left on device解决方法，在宿主机执行：sudo sysctl fs.inotify.max_user_watches=1048576

1. 安装docker-compose参考https://www.huaweicloud.com/kunpeng/software/dockercompose.html最后，将docker-compose-Linux-aarch64拷贝到/usr/bin目录下2. 在/etc/profile中加入如下内容export GOPATH="/home/xxx/go"export P...

DaemonSet、Deployment、StatefulSet 和 ReplicaSet 在 v1.16 中将不再从 extensions/v1beta1、apps/v1beta1 或 apps/v1beta2 提供服务解决方法是：将yml配置文件内的api接口修改为 apps/v1 ，导致原因为之间使用的kubernetes 版本是1.14.x版本，1.16.x 版本放弃部分API支持...

在arm64平台下，执行：gem install zookeeper报错：configure: error: cannot guess build type; you must specify one说明，/var/lib/gems/2.3.0/gems/zookeeper-1.4.11/ext/zkc-3.4.5/c/config.sub文件已经过时，没有包含当前arm64...