Hey, In addition to what Alex said (both of us are in the Tectonic Installer team):
For an automated bootstrap process of k8s components you need to address two problems besides the infrastructure provisioning part: 1. executing the deployment on exactly one bootstrap node. We call this internally "bootstrap master election". This can be solved on: a) the machine initiating the automated installation by choosing one machine and simply executing a remote command a la `ssh user@bootstrap-master bootstrap.sh` where `bootstrap.sh` includes your `kubectl apply's`. We use terraform to do this for us, see [1]. b) if the machine initiating the automated installation provisions the infrastructure, but does not have access to the actual nodes, the election process is more complex. On AWS for instance we have a small script being executed on each master which detects whether it is meant to be a bootstrap node, see [2]. Obviously the detection mechanism depends on the platform. 2. at-most-once execution on the bootstrap node itself. We solved this using a simple systemd trick, see [3] and [4] ;-) I hope that helps a bit. - Sergiusz [1] https://github.com/coreos/tectonic-installer/blob/5c4fd74/platforms/metal/remote.tf#L73-L100 [2] https://github.com/coreos/tectonic-installer/blob/5c4fd74/modules/aws/ignition/resources/detect-master.sh [3] https://github.com/coreos/tectonic-installer/blob/5c4fd74/modules/bootkube/resources/bootkube.service#L3 [4] https://www.freedesktop.org/software/systemd/man/systemd.unit.html#ConditionArchitecture= On Wednesday, August 16, 2017 at 3:52:59 AM UTC+2, Marcelo Andrade wrote: > Hey folks! > > I'm trying to put together some automation on the process of creating our > Kubernetes clusters and I got all the infrastructure/OS/master/nodes basic > installation covered. > > > Then I figured I have no convenient way to automatize the launch of our > several services that already run under Kubernetes > > > Is there a smart to "bootstrap" kubernetes elements on the startup of the > cluster, let's say tthe DNS addon (creating the SA, configmap, > deployment...), the ingress-controller components (deployment for > default-backend, configmaps and so) or whatever client-controlled deployment? > > > I could work around scripting some kubectls or posts on the API but then > ocurred 'there must be an easier way"... Or there is not? :D -- You received this message because you are subscribed to the Google Groups "Kubernetes user discussion and Q&A" group. To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-users+unsubscr...@googlegroups.com. To post to this group, send email to kubernetes-users@googlegroups.com. Visit this group at https://groups.google.com/group/kubernetes-users. For more options, visit https://groups.google.com/d/optout.
