* Observed frequent kube-scheduler and controller-manager
restarts with Calico as the CNI provider. Root cause was
unclear since control plane was functional and tests of
pod to pod network connectivity passed
* Root cause: Calico sets up cali* and tunl* network interfaces
for containers on hosts. NetworkManager tries to manage these
interfaces. It periodically disconnected veth pairs. Logs did
not surface this issue since its not an error per-se, just Calico
and NetworkManager dueling for control. Kubernetes correctly
restarted pods failing health checks and ensured 2 replicas were
running so the control plane functioned mostly normally. Pod to
pod connecitivity was only affected occassionally. Pain to debug.
* Solution: Configure NetworkManager to ignore the Calico ifaces
per Calico's recommendation. Cloud-init writes files after
NetworkManager starts, so a restart is required on first boot. On
subsequent boots, the file is present so no restart is needed
* (containerized) kube-proxy warns that it is unable to
load the ip_vs kernel module despite having the correct
mounts. Atomic uses an xz compressed module and modprobe
in the container was not compiled with compression support
* Workaround issue for now by always loading ip_vs on-host
* https://github.com/kubernetes/kubernetes/issues/60
* Use the upstream bootkube image packaged with the
required metadata to be usable as a system container
under systemd
* Run bootkube with runc so no host level components
use Docker any more. Docker is still the runtime
* Remove bootkube script and old systemd unit
* Change kubelet system image to use --cgroups-per-qos=true
(default) instead of false
* Change kubelet system image to use --enforce-node-allocatable=pods
instead of an empty string
* Fix kubelet port-forward on Google Cloud / Fedora Atomic
* Mount the host's /etc/hosts in kubelet system containers
* Problem: kubelet runc system containers on Atomic were not
mounting the host's /etc/hosts, like rkt-fly does on Container
Linux. `kubectl port-forward` calls socat with localhost. DNS
servers on AWS, DO, and in many bare-metal environments resolve
localhost to the caller as a convenience. Google Cloud notably
does not nor is it required to do so and this surfaced the
missing /etc/hosts in runc kubelet namespaces.
* Enable etcd v3.3 metrics to expose metrics for
scraping by Prometheus
* Use k8s.gcr.io instead of gcr.io/google_containers
* Add flexvolume plugin mount to controller manager
* Update kube-dns from v1.14.8 to v1.14.9
* Mount /opt/cni/bin in kubelet system container so
CNI plugin binaries can be found. Before, flannel
worked because the kubelet falls back to flannel
plugin baked into the hyperkube (undesired)
* Move the CNI bin install location later, since /opt
changes may be lost between ostree rebases
* Use the upstream hyperkube image packaged with the
required metadata to be usable as a system container
under systemd
* Fix port-forward since socat is included
* Several known hacks and broken areas
* Download v1.10 Kubelet from release tarball
* Install flannel CNI binaries to /opt/cni
* Switch SELinux to Permissive
* Disable firewalld service
* port-forward won't work, socat missing
* Terraform v0.11.4 introduced changes to remote-exec
that mean Typhoon bare-metal clusters require multiple
runs of terraform apply to ssh and bootstrap.
* Bare-metal installs PXE boot a live instance to install
to disk and then reboot from disk as controllers/workers.
Terraform remote-exec has no way to "know" to wait until
the reboot has occurred to kickoff Kubernetes bootstrap.
Previously Typhoon created a "debug" user during this
install phase to allow an admin to SSH, but remote-exec
would hang, trying to connect as user "core". Terraform
v0.11.4 changes this behavior so remote-exec fails and
a user must re-run terraform apply until succeeding.
* A new way to "trick" remote-exec into waiting for the
reboot into the disk install is to run SSH on a non-standard
port during the disk install. This retains the ability
for an admin to SSH during install (most distros don't have
this) and fixes the issue so only a single run of terraform
apply is needed.
* https://github.com/hashicorp/terraform/pull/17359#issuecomment-376415464
* Use etcd v3.3 --listen-metrics-urls to expose only metrics
data via http://0.0.0.0:2381 on controllers
* Add Prometheus discovery for etcd peers on controller nodes
* Temporarily drop two noisy Prometheus alerts
* Add a node-role.kubernetes.io/controller="true" node label
to controllers so Prometheus service discovery can filter to
services that only run on controllers (i.e. masters)
* Leave node-role.kubernetes.io/master="" untouched as its
a Kubernetes convention
* Remove PersistentVolumeLabel admission controller flag
* Switch Deployments and DaemonSets to apps/v1
* Minor update to pod-checkpointer image version
* Add flannel service account and limited RBAC cluster role
* Change DaemonSets to tolerate NoSchedule and NoExecute taints
* Remove deprecated apiserver --etcd-quorum-read flag
* Update Calico from v3.0.1 to v3.0.2
* Add Calico GlobalNetworkSet CRD
* https://github.com/poseidon/terraform-render-bootkube/pull/44
* Create separate container-linux-install profiles (and
cached-container-linux-install) for each node in a cluster
* Fix contention bug on bare-metal during `terraform apply`.
With only a global install profile, terraform would create
(or retain) the profile for each cluster and try to delete
it for each cluster being deleted. As a result, in some cases
apply had to be run multiple times before terraform's repr
of constraints was satisfied (profile deleted and recreated)
* Allow Container Linux install properties to vary between
clusters, such as using a different Container Linux channel
or version for different clusters
* Allow kube-dns to respond to DNS queries with a custom
suffix, instead of the default 'cluster.local'
* Useful when multiple clusters exist on the same local
network and wish to query services on one another
* When restarting masters, `etcd-member.service` may fail to lookup peers if
/etc/resolv.conf hasn't been populated yet. Require the wait-for-dns.service.
* Update hyperkube from v1.8.3 to v1.8.4
* Remove flock from bootstrap-apiserver and kube-apiserver
* Remove unused critical-pod annotations in manifests
* Use service accounts for kube-proxy and pod-checkpointer
* Update Calico from v2.6.1 to v2.6.3
* Update flannel from v0.9.0 to v0.9.1
* Remove Calico termination grace period to prevent calico
from getting stuck for extended periods
* https://github.com/poseidon/terraform-render-bootkube/pull/29
* Kubernetes v1.8.2 fixes a memory leak in the v1.8.1 apiserver
* Switch to using the `gcr.io/google_containers/hyperkube` for the
on-host kubelet and shutdown drains
* Update terraform-render-bootkube manifests generation
* Update flannel from v0.8.0 to v0.9.0
* Add `hairpinMode` to flannel CNI config
* Add `--no-negcache` to kube-dns dnsmasq
* Without this change, if a cluster doesn't set the controller
or worker networkd lists, an err "element() may not be used
with an empty list" occurs.
* controller_networkds and worker_networks are intended to be
optional and temporary, not required at all
* Accept ordered lists of controller and worker networkd configs
* Do not rely on these variables. They will be replaced with a
cleaner mechanism at a future date
* Mounting NFS exports as volumes from some NFS servers fails because
the kubelet isn't starting rpc-statd as expected. Describing pods
that are stuck creating shows rpc.statd is required for remote locking
* Starting rpc-statd.service resolves the issue and all NFS mounts
seem to be working.
* Recommended approach https://github.com/coreos/bugs/issues/2074
* Template bare-metal Container Linux configs with Terraform's
(limited) template_file module. This allows rendering problems
to be identified during `terraform plan` and is favored over
using the Matchbox templating feature when the configs are
served to PXE booting nodes.
* Writes a Matchbox profile for each machine, which will be served
as-is. The effect is the same, each node gets provisioned with its
own Container Linux config.
