Changed loglevel from error to warning for missing disable_ipv6 file. docker/libnetwork#2223
Fixed subnet allocation to avoid reallocating recently freed subnets. docker/libnetwork#2255
Fixed libnetwork issue which caused errors to be returned when iptables or firewalld issues transient warnings. docker/libnetwork#2218
Fixed too many “Plugin not found” error messages. moby/moby#36119
Added failed allocations retry immediately upon a deallocation to overcome IP exhaustion. docker/swarmkit#2711
Fixed leaking task resources. docker/swarmkit#2755
Fixed deadlock in dispatcher that could cause node to crash. docker/swarmkit#2753
When all Swarm managers are stopped at the same time, the swarm might end up in a
split-brain scenario. Learn more.
Under certain conditions, swarm leader re-election may timeout
prematurely. During this period, docker commands may fail. Also during
this time, creation of globally-scoped networks may be unstable. As a
workaround, wait for leader election to complete before issuing commands
to the cluster.
It’s recommended that users create overlay networks with /24 blocks (the default) of 256 IP addresses when networks are used by services created using VIP-based endpoint-mode (the default). This is because of limitations with Docker Swarm moby/moby#30820. Users should not work around this by increasing the IP block size. To work around this limitation, either use dnsrr endpoint-mode or use multiple smaller overlay networks.
Docker may experience IP exhaustion if many tasks are assigned to a single overlay network, for example if many services are attached to that network or because services on the network are scaled to many replicas. The problem may also manifest when tasks are rescheduled because of node failures. In case of node failure, Docker currently waits 24h to release overlay IP addresses. The problem can be diagnosed by looking for failed to allocate network IP for task messages in the Docker logs.
SELinux enablement is not supported for containers on IBM Z on RHEL because of missing Red Hat package.
If a container is spawned on node A, using the same IP of a container destroyed
on nodeB within 5 min from the time that it exit, the container on node A is
not reachable until one of these 2 conditions happens:
Container on A sends a packet out,
The timer that cleans the arp entry in the overlay namespace is triggered (around 5 minutes).
As a workaround, send at least a packet out from each container like
(ping, GARP, etc).
Docker 是一个开源的应用容器引擎，让开发者可以打包他们的应用以及依赖包到一个可移植的镜像中，然后发布到任何流行的 Linux或Windows 机器上，也可以实现虚拟化。容器是完全使用沙箱机制，相互之间不会有任何接口。