我最近通过https://github.com/tarosky/k8s-redis-ha设置了redis，这个repo包含一个init容器，我已经包含了一个额外的init容器，以便设置密码等。
我看到一些奇怪的（并且似乎没有记录）行为，其中init容器在redis容器启动之前按预期运行，然后它们随后每小时运行，接近一小时。我已经使用busybox init容器（它什么都不做）在部署和测试中测试了这种行为。 statefulset并体验相同的行为，因此它不是特定于此redis pod。
我在裸机上使用k8s 1.6和1.8进行了测试，结果相同，但是当将初始容器应用于GKE（k8s 1.7）时，这种情况不会发生。我无法看到GKE kubelet指示此行为的任何标记。
请参阅下面的kubectl describe pod，显示当主pod未退出/崩溃时运行init容器。
Name: redis-sentinel-1
Namespace: (redacted)
Node: (redacted)/(redacted)
Start Time: Mon, 12 Mar 2018 06:20:55 +0000
Labels: app=redis-sentinel
controller-revision-hash=redis-sentinel-7cc557cf7c
Annotations: kubernetes.io/created-by={"kind":"SerializedReference","apiVersion":"v1","reference":{"kind":"StatefulSet","namespace":"(redacted)","name":"redis-sentinel","uid":"759a3a3b-25bd-11e8-a8ce-0242ac110...
security.alpha.kubernetes.io/unsafe-sysctls=net.core.somaxconn=1024
Status: Running
IP: (redacted)
Controllers: StatefulSet/redis-sentinel
Init Containers:
redis-ha-server:
Container ID: docker://557d777a7c660b062662426ebe9bbf6f9725fb9d88f89615a8881346587c1835
Image: tarosky/k8s-redis-ha:sentinel-3.0.1
Image ID: docker-pullable://tarosky/k8s-redis-ha@sha256:98e09ef5fbea5bfd2eb1858775c967fa86a92df48e2ec5d0b405f7ca3f5ada1c
Port:
State: Terminated
Reason: Completed
Exit Code: 0
Started: Tue, 13 Mar 2018 03:01:12 +0000
Finished: Tue, 13 Mar 2018 03:01:12 +0000
Ready: True
Restart Count: 0
Environment: <none>
Mounts:
/opt from opt (rw)
/var/run/secrets/kubernetes.io/serviceaccount from default-token-hkj6d (ro)
-redis-init:
Container ID: docker://18c4e353233a6827999ae4a16adf1f408754a21d80a8e3374750fdf9b54f9b1a
Image: gcr.io/(redacted)/redis-init
Image ID: docker-pullable://gcr.io/(redacted)/redis-init@sha256:42042093d58aa597cce4397148a2f1c7967db689256ed4cc8d9f42b34d53aca2
Port:
State: Terminated
Reason: Completed
Exit Code: 0
Started: Tue, 13 Mar 2018 03:01:25 +0000
Finished: Tue, 13 Mar 2018 03:01:25 +0000
Ready: True
Restart Count: 0
Environment: <none>
Mounts:
/opt from opt (rw)
/secrets/redis-password from redis-password (rw)
/var/run/secrets/kubernetes.io/serviceaccount from default-token-hkj6d (ro)
Containers:
redis-sentinel:
Container ID: docker://a54048cbb7ec535c841022c543a0d566c9327f37ede3a6232516721f0e37404d
Image: redis:3.2
Image ID: docker-pullable://redis@sha256:474fb41b08bcebc933c6337a7db1dc7131380ee29b7a1b64a7ab71dad03ad718
Port: 26379/TCP
Command:
/opt/bin/k8s-redis-ha-sentinel
Args:
/opt/sentinel.conf
State: Running
Started: Mon, 12 Mar 2018 06:21:02 +0000
Ready: True
Restart Count: 0
Readiness: exec [redis-cli -p 26379 info server] delay=0s timeout=1s period=10s #success=1 #failure=3
Environment:
SERVICE: redis-server
SERVICE_PORT: redis-server
Mounts:
/opt from opt (rw)
/var/run/secrets/kubernetes.io/serviceaccount from default-token-hkj6d (ro)
redis-sword:
Container ID: docker://50279448bbbf175b6f56f96dab59061c4652c2117452ed15b3a5380681c7176f
Image: tarosky/k8s-redis-ha:sword-3.0.1
Image ID: docker-pullable://tarosky/k8s-redis-ha@sha256:2315c7a47d9e47043d030da270c9a1252c2cfe29c6e381c8f50ca41d3065db6d
Port:
State: Running
Started: Mon, 12 Mar 2018 06:21:03 +0000
Ready: True
Restart Count: 0
Environment:
SERVICE: redis-server
SERVICE_PORT: redis-server
SENTINEL: redis-sentinel
SENTINEL_PORT: redis-sentinel
Mounts:
/opt from opt (rw)
/var/run/secrets/kubernetes.io/serviceaccount from default-token-hkj6d (ro)
Conditions:
Type Status
Initialized True
Ready True
PodScheduled True
Volumes:
opt:
Type: HostPath (bare host directory volume)
Path: /store/redis-sentinel/opt
redis-password:
Type: Secret (a volume populated by a Secret)
SecretName: redis-password
Optional: false
default-token-hkj6d:
Type: Secret (a volume populated by a Secret)
SecretName: default-token-hkj6d
Optional: false
QoS Class: BestEffort
Node-Selectors: <none>
Tolerations: <none>
Events:
FirstSeen LastSeen Count From SubObjectPath Type Reason Message
--------- -------- ----- ---- ------------- -------- ------ -------
20h 30m 21 kubelet, 10.1.3.102 spec.initContainers{redis-ha-server} Normal Pulling pulling image "tarosky/k8s-redis-ha:sentinel-3.0.1"
21h 30m 22 kubelet, 10.1.3.102 spec.initContainers{redis-ha-server} Normal Started Started container
21h 30m 22 kubelet, 10.1.3.102 spec.initContainers{redis-ha-server} Normal Created Created container
20h 30m 21 kubelet, 10.1.3.102 spec.initContainers{redis-ha-server} Normal Pulled Successfully pulled image "tarosky/k8s-redis-ha:sentinel-3.0.1"
21h 30m 22 kubelet, 10.1.3.102 spec.initContainers{redis-init} Normal Pulling pulling image "gcr.io/(redacted)/redis-init"
21h 30m 22 kubelet, 10.1.3.102 spec.initContainers{redis-init} Normal Pulled Successfully pulled image "gcr.io/(redacted)/redis-init"
21h 30m 22 kubelet, 10.1.3.102 spec.initContainers{redis-init} Normal Created Created container
21h 30m 22 kubelet, 10.1.3.102 spec.initContainers{redis-init} Normal Started Started container
请注意从Mon, 12 Mar 2018 06:21:02 +0000（0重启）开始的pod中的容器以及从Tue, 13 Mar 2018 03:01:12 +0000开始的Init Containers。这些似乎每隔一小时就会重新运行一次。
我们的裸机必须在某处错误配置init容器？任何人都可以对这种奇怪的行为有所了解吗？
答案 0 :(得分：0)
如果要修剪掉已退出的容器，则可能是原因是修剪/移除容器。在我的测试中，似乎退出的初始化容器（每小时或以其他方式）已从Docker引擎中删除，例如使用“ docker system prune -f”，将导致Kubernetes重新启动初始化容器。如果仍然存在，这是您的问题吗？
另外，请参阅https://kubernetes.io/docs/concepts/cluster-administration/kubelet-garbage-collection/，以获取Kubelet垃圾收集文档，该文档似乎支持这些类型的任务（而不是需要自己实现）