given that etcd is disabled in the nodes we need to wait for master's etcd to be
available so that flanneld gets properly wired.
Signed-off-by: António Meireles <antonio.meireles@reformi.st>
avoids race conditions (docker/docker-registry#892) when multiple workers are
attempting to run sql migrations simultaneously. dunno why this isn't the
registry's default since this isn't specially hard to trigger so we're just
playing safe here.
Signed-off-by: António Meireles <antonio.meireles@reformi.st>
in the footsteps of #5254 merge earlier today and per @brendandburns comments
in the PR review thread...
Signed-off-by: António Meireles <antonio.meireles@reformi.st>
a more idiomatic take on #4378. as a bonus, and to be nice on
resources, an ultra lightweight docker-registry container is
consumed.
Signed-off-by: António Meireles <antonio.meireles@reformi.st>
Update node.yaml to support graphdriver: overlay as btrfs produces "fatal" "prerequisites for driver not satisfied (wrong filesystem?)" on CoreOS Alpha 561.0.0
* CoreOS guides now use the 0.5.4 official release binaries
* kube-register has been updated to work with the 0.5.4 release
* flannel uses the vxlan backend by default for better performance
* cloud-configs are used for all examples
* instructions for GCE, AWS, and VMware Fusion have been combined