昨天晚上搭建好了k8s多主集群,启动了一个nginx的pod,然而每启动一个pod就伴随这一个pause容器,考虑到之前在做kubelet的systemd unit文件时有见到:
[root@node01 ~]# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
43884d79fe6f nginx "nginx -g 'daemon of…" 11 hours ago Up 11 hours k8s_nginx-pod_nginx-con-594b8d6b48-46gf7_default_25b0048f-a24f-11e9-8149-00163e0134cf_0
eff67394c9c8 nginx "nginx -g 'daemon of…" 11 hours ago Up 11 hours k8s_nginx-pod_nginx-con-594b8d6b48-vt589_default_25aefc99-a24f-11e9-8149-00163e0134cf_1
261226f6b92a registry.cn-hangzhou.aliyuncs.com/google_containers/pause-amd64:3.1 "/pause" 11 hours ago Up 11 hours k8s_POD_nginx-con-594b8d6b48-vt589_default_25aefc99-a24f-11e9-8149-00163e0134cf_1
fc94013b93dd registry.cn-hangzhou.aliyuncs.com/google_containers/pause-amd64:3.1 "/pause" 11 hours ago Up 11 hours k8s_POD_nginx-con-594b8d6b48-46gf7_default_25b0048f-a24f-11e9-8149-00163e0134cf_1
### 6 创建和分发kubelet systemd unit文件
```bash
cat >/etc/systemd/system/kubelet.service <<EOF
[Unit]
Description=Kubernetes Kubelet
Documentation=https://github.com/GoogleCloudPlatform/kubernetes
After=docker.service
Requires=docker.service
[Service]
WorkingDirectory=/var/lib/kubelet
ExecStart=/usr/local/bin/kubelet \
--bootstrap-kubeconfig=/etc/kubernetes/cert/kubelet-bootstrap.kubeconfig \
--cert-dir=/etc/kubernetes/cert \
--kubeconfig=/etc/kubernetes/cert/kubelet.kubeconfig \
--config=/etc/kubernetes/cert/kubelet.config.json \
--hostname-override=172.24.150.89 \
--pod-infra-container-image=registry.cn-hangzhou.aliyuncs.com/google_containers/pause-amd64:3.1 \
--allow-privileged=true \
--alsologtostderr=true \
--logtostderr=false \
--log-dir=/var/log/kubernetes \
--v=2
Restart=on-failure
RestartSec=5
[Install]
WantedBy=multi-user.target
EOF
首先找一下源码看一下:
pause源码C语言编写的主要有四个文件:
orphan.c文件
#include <stdio.h>
#include <unistd.h>
int main() {
pid_t pid;
pid = fork();
if (pid == 0) {
while (getppid() > 1)
;
printf("Child exiting: pid=%d ppid=%d\n", getpid(), getppid());
return 0;
} else if (pid > 0) {
printf("Parent exiting: pid=%d ppid=%d\n", getpid(), getppid());
return 0;
}
perror("Could not create child");
return 1;
}
pause.c文件
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <unistd.h>
#define STRINGIFY(x) #x
#define VERSION_STRING(x) STRINGIFY(x)
#ifndef VERSION
#define VERSION HEAD
#endif
static void sigdown(int signo) {
psignal(signo, "Shutting down, got signal");
exit(0);
}
static void sigreap(int signo) {
while (waitpid(-1, NULL, WNOHANG) > 0)
;
}
int main(int argc, char **argv) {
int i;
for (i = 1; i < argc; ++i) {
if (!strcasecmp(argv[i], "-v")) {
printf("pause.c %s\n", VERSION_STRING(VERSION));
return 0;
}
}
if (getpid() != 1)
/* Not an error because pause sees use outside of infra containers. */
fprintf(stderr, "Warning: pause should be the first process\n");
if (sigaction(SIGINT, &(struct sigaction){.sa_handler = sigdown}, NULL) < 0)
return 1;
if (sigaction(SIGTERM, &(struct sigaction){.sa_handler = sigdown}, NULL) < 0)
return 2;
if (sigaction(SIGCHLD, &(struct sigaction){.sa_handler = sigreap,
.sa_flags = SA_NOCLDSTOP},
NULL) < 0)
return 3;
for (;;)
pause();
fprintf(stderr, "Error: infinite loop terminated\n");
return 42;
}
至于里面写的啥,我也看不懂,pose出来让你们看看哈!
Makefile文件用于制作pause镜像,制作镜像的模板是Dockerfile,首先看这个Dockerfile(去除注释)文件:
FROM scratch
ARG ARCH
ADD bin/pause-${ARCH} /pause
ENTRYPOINT ["/pause"]
FROM scratch: 基础镜像是一个空镜像(an explicity empty image)
ARG ARCH: 等待在docker-build -build-arg时提供的ARCH参数
ADD bin/pause-$(ARCH)/pause: 添加外部文件到内部
ENTRYPOINT[“/pause”]: 开启容器,运行命令
中间两部非常重要,至于如何实现的不懂哦?
pause的Makefile
.PHONY: all push container clean orphan all-push push-manifest
include ../../hack/make-rules/Makefile.manifest
REGISTRY ?= staging-k8s.gcr.io
IMAGE = $(REGISTRY)/pause
IMAGE_WITH_ARCH = $(IMAGE)-$(ARCH)
TAG = 3.1
REV = $(shell git describe --contains --always --match='v*')
# Architectures supported: amd64, arm, arm64, ppc64le and s390x
ARCH ?= amd64
ALL_ARCH = amd64 arm arm64 ppc64le s390x
CFLAGS = -Os -Wall -Werror -static -DVERSION=v$(TAG)-$(REV)
KUBE_CROSS_IMAGE ?= k8s.gcr.io/kube-cross
KUBE_CROSS_VERSION ?= $(shell cat ../build-image/cross/VERSION)
BIN = pause
SRCS = pause.c
ifeq ($(ARCH),amd64)
TRIPLE ?= x86_64-linux-gnu
endif
ifeq ($(ARCH),arm)
TRIPLE ?= arm-linux-gnueabihf
endif
ifeq ($(ARCH),arm64)
TRIPLE ?= aarch64-linux-gnu
endif
ifeq ($(ARCH),ppc64le)
TRIPLE ?= powerpc64le-linux-gnu
endif
ifeq ($(ARCH),s390x)
TRIPLE ?= s390x-linux-gnu
endif
# If you want to build AND push all containers, see the 'all-push' rule.
all: all-container
all-push: all-push-images push-manifest
push-manifest: manifest-tool
manifest-tool push from-args --platforms $(call join_platforms,$(ALL_ARCH)) --template $(IMAGE)-ARCH:$(TAG) --target $(IMAGE):$(TAG)
sub-container-%:
$(MAKE) ARCH=$* container
sub-push-%:
$(MAKE) ARCH=$* push
all-container: $(addprefix sub-container-,$(ALL_ARCH))
all-push-images: $(addprefix sub-push-,$(ALL_ARCH))
build: bin/$(BIN)-$(ARCH)
# 先启动一个容器
```bash
bin/$(BIN)-$(ARCH): $(SRCS)
mkdir -p bin
docker run --rm -u $$(id -u):$$(id -g) -v $$(pwd):/build \
$(KUBE_CROSS_IMAGE):$(KUBE_CROSS_VERSION) \
/bin/bash -c "\
cd /build && \
$(TRIPLE)-gcc $(CFLAGS) -o $@ $^ && \
$(TRIPLE)-strip $@"
根据上面的启动容器,来build一个pause镜像
container: .container-$(ARCH)
.container-$(ARCH): bin/$(BIN)-$(ARCH)
docker build --pull -t $(IMAGE_WITH_ARCH):$(TAG) --build-arg ARCH=$(ARCH) .
touch $@
push: .push-$(ARCH)
.push-$(ARCH): .container-$(ARCH)
docker push $(IMAGE_WITH_ARCH):$(TAG)
touch $@
# Useful for testing, not automatically included in container image
orphan: bin/orphan-$(ARCH)
bin/orphan-$(ARCH): orphan.c
mkdir -p bin
docker run -u $$(id -u):$$(id -g) -v $$(pwd):/build \
$(KUBE_CROSS_IMAGE):$(KUBE_CROSS_VERSION) \
/bin/bash -c "\
cd /build && \
$(TRIPLE)-gcc $(CFLAGS) -o $@ $^ && \
$(TRIPLE)-strip $@"
clean:
rm -rf .container-* .push-* bin/
ARCH值:
# Architectures supported: amd64, arm, arm64, ppc64le and s390x
ARCH ?= amd64
ALL_ARCH = amd64 arm arm64 ppc64le s390x
可以看出支持很多架构类型,默认为amd64
制作pause镜像的过程如下:
REGISTRY ?= staging-k8s.gcr.io
IMAGE = $(REGISTRY)/pause
IMAGE_WITH_ARCH = $(IMAGE)-$(ARCH)
TAG = 3.1
ARCH ?= amd64
BIN = pause
SRCS = pause.c
REV = $(shell git describe --contains --always --match='v*')
CFLAGS = -Os -Wall -Werror -static -DVERSION=v$(TAG)-$(REV)
KUBE_CROSS_IMAGE ?= k8s.gcr.io/kube-cross
KUBE_CROSS_VERSION ?= $(shell cat ../build-image/cross/VERSION)
ifeq ($(ARCH),amd64)
TRIPLE ?= x86_64-linux-gnu
build: bin/$(BIN)-$(ARCH)
# 启动一个容器
bin/$(BIN)-$(ARCH): $(SRCS)
mkdir -p bin
docker run --rm -u $$(id -u):$$(id -g) -v $$(pwd):/build \
$(KUBE_CROSS_IMAGE):$(KUBE_CROSS_VERSION) \
/bin/bash -c "\
cd /build && \
$(TRIPLE)-gcc $(CFLAGS) -o $@ $^ && \
$(TRIPLE)-strip $@"
# build一个pause镜像
container: .container-$(ARCH)
.container-$(ARCH): bin/$(BIN)-$(ARCH)
docker build --pull -t $(IMAGE_WITH_ARCH):$(TAG) --build-arg ARCH=$(ARCH) .
touch $@
$(KUBE_CROSS_IMAGE):$(KUBE_CROSS_VERSION)=k8s.gcr.io/kube-cross:v1.12.6-1
启动一个容器
bin/pause-amd64:pause.c
mkdir -p bin
docker run --rm -u $$(id -u):$$(id -g) -v $$(pwd):/build \
k8s.gcr.io/kube-cross:v1.12.6-1 \
/bin/bash -c "\
cd /build && \
x86_64-linux-gnu-gcc v3.1-$(REV) -o $@ $^ && \
x86_64-linux-gnu-strip $@"
# build一个pause镜像
container: .container-amd64
.container-amd64: bin/pause-amd64
docker build --pull -t staging-k8s.gcr.io/pause-amd64:3.1 --build-arg ARCH=amd64 .
至于详细的构建过程,我就不写了, 我看不懂呀!
pause容器的工作
可知kubernetes的pod抽象基于Linux的namespace和cgroups,为容器提供了良好的隔离环境。在同一个pod中,不同容器犹如在localhost中。
在Unix系统中,PID为1的进程为init进程,即所有进程的父进程。它很特殊,维护一张进程表,不断地检查进程状态。例如,一旦某个子进程由于父进程的错误而变成了“孤儿进程”,其便会被init进程进行收养并最终回收资源,从而结束进程。
或者,某子进程已经停止但进程表中仍然存在该进程,因为其父进程未进行wait syscall进行索引,从而该进程变成“僵尸进程”,这种僵尸进程存在时间较短。不过如果父进程只wait,而未syscall的话,僵尸进程便会存在较长时间。
同时,init进程不能处理某个信号逻辑,拥有“信号屏蔽”功能,从而防止init进程被误杀。
容器中使用pid namespace来对pid进行隔离,从而每个容器中均有其独立的init进程。例如对于寄主机上可以用个发送SIGKILL或者SIGSTOP(也就是docker kill 或者docker stop)来强制终止容器的运行,即终止容器内的init进程。一旦init进程被销毁, 同一pid namespace下的进程也随之被销毁,并容器进程被回收相应资源。
kubernetes中的pause容器便被设计成为每个业务容器提供以下功能:
在pod中担任Linux命名空间共享的基础;
启用pid命名空间,开启init进程。
[root@harbor1 ~]# docker run -idt --name pause-ceshi cntsp/pause:3.1
ce2a4ffccf1d2b190b9f8cc45f6e4912f278aa7a55280ce8a69433e5345e5816
[root@harbor1 ~]# docker run -idt --name busybox11 --net=container:pause11 --pid=container:pause11 --ipc=container:pause11 busybox
docker: Error response from daemon: No such container: pause11.
See 'docker run --help'.
[root@harbor1 ~]# docker run -idt --name busybox11 --net=container:pause-ceshi --pid=container:pause-ceshi --ipc=container:pause-ceshi busybox
Unable to find image 'busybox:latest' locally
latest: Pulling from library/busybox
8e674ad76dce: Pull complete
Digest: sha256:c94cf1b87ccb80f2e6414ef913c748b105060debda482058d2b8d0fce39f11b9
Status: Downloaded newer image for busybox:latest
6bec8c69198eaa444580ccaa0a50afbc16a51eea489fa7e81729d535cb691ea6
[root@harbor1 ~]# docker ps -a
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
6bec8c69198e busybox "sh" bout a minute ago Up About a minute busybox11
ce2a4ffccf1d cntsp/pause:3.1 "/pause" 2 minutes ago Up 2 minutes pause-ceshi
[root@harbor1 ~]# docker exec -it busybox11 /bin/bash
OCI runtime exec failed: exec failed: container_linux.go:348: starting container process caused "exec: \"/bin/bash\": stat /bin/bash: no such file or directory": unknown
[root@harbor1 ~]# docker exec -it busybox11 /bin/sh
/ # ps aux
PID USER TIME COMMAND
1 root 0:00 /pause
6 root 0:00 sh
16 root 0:00 /bin/sh
21 root 0:00 ps aux
/ #
这里补充一点:Docker网络模式:
其中有一种是Container:
> * 复用其它容器的网络
Container模式:
> * 重用另一个容器的Network Namespace,新创建的容器不会创建网卡,配置IP
> * 和一个指定的容器共享IP、端口范围
> * 两个容器除了网络方面,其它的如文件系统、进程列表等还是隔离的,两个容器的进程可以通过lo网卡设备通信。
> * Pod里的Pause容器和Pod里的其它容器就是此种网络