Setting the following:
```
kube_kubeadm_controller_extra_args:
address: 0.0.0.0
terminated-pod-gc-threshold: "100"
```
Results in `terminated-pod-gc-threshold: 100` in the kubeadm config file. But it has to be a string to work.
to the API server configuration.
This solves the problem where if you have non-resolvable node names,
and try to scale the server by adding new nodes, kubectl commands
start to fail for newly added nodes, giving a TCP timeout error when
trying to resolve the node hostname against a public DNS.
* Fix run kubectl error
Fix run kubectl error when first master doesn't work
* if access_ip is define use first_kube_master
else different master use a different ip
* Delete set first_kube_master and use kube_apiserver_access_address
* Set filemode to 0640
weave-net.yml file is readable by all users on the host. It however contains the weave_password to encrypt all pod communication. It should only be readable by root.
* Set mode 0640 on users_file with basic auth
* Added cilium support
* Fix typo in debian test config
* Remove empty lines
* Changed cilium version from <latest> to <v1.0.0-rc3>
* Add missing changes for cilium
* Add cilium to CI pipeline
* Fix wrong file name
* Check kernel version for cilium
* fixed ci error
* fixed cilium-ds.j2 template
* added waiting for cilium pods to run
* Fixed missing EOF
* Fixed trailing spaces
* Fixed trailing spaces
* Fixed trailing spaces
* Fixed too many blank lines
* Updated tolerations,annotations in cilium DS template
* Set cilium_version to iptables-1.9 to see if bug is fixed in CI
* Update cilium image tag to v1.0.0-rc4
* Update Cilium test case CI vars filenames
* Add optional prometheus flag, adjust initial readiness delay
* Update README.md with cilium info
Even though there it kubeadm_token_ttl=0 which means that kubeadm token never expires, it is not present in `kubeadm token list` after cluster is provisioned (at least after it is running for some time) and there is issue regarding this https://github.com/kubernetes/kubeadm/issues/335, so we need to create a new temporary token during the cluster upgrade.
Ansible automatically installs the python-apt package when using
the 'apt' Ansible module, if python-apt is not present. This patch
removes the (unneeded) explicit installation in the Kubespray
'preinstall' role.
* allow installs to not have hostname overriden with fqdn from inventory
* calico-config no longer requires local as and will default to global
* when cloudprovider is not defined, use the inventory_hostname for cni-calico
* allow reset to not restart network (buggy nodes die with this cmd)
* default kube_override_hostname to inventory_hostname instead of ansible_hostname
Cloud resolvers are mandatory for hosts on GCE and OpenStack
clouds. The 8.8.8.8 alternative resolver was dropped because
there is already a default nameserver. The new var name
reflects the purpose better.
Also restart apiserver when modifying dns settings.
If you configure your external loadbalancer to do a simple tcp pass-through to the api servers, and you do not use a DNS FQDN but just the ip, then you need to add the ip adress to the certificates too.
Example config:
```
## External LB example config
apiserver_loadbalancer_domain_name: "10.50.63.10"
loadbalancer_apiserver:
address: 10.50.63.10
port: 8383
```
kube-proxy is complaining of missing modules at startup. There is a plan
to also support an LVS implementation of kube-proxy in additon to
userspace and iptables
Update checksum for kubeadm
Use v1.9.0 kubeadm params
Include hash of ca.crt for kubeadm join
Update tag for testing upgrades
Add workaround for testing upgrades
Remove scale CI scenarios because of slow inventory parsing
in ansible 2.4.x.
Change region for tests to us-central1 to
improve ansible performance
Starting with Kubernetes v1.8.4, kubelet ignores the AWS cloud
provider string and uses the override hostname, which fails
Node admission checks.
Fixes#2094
As we have seen with other containers, sometimes container removal fails on the first attempt due to some Docker bugs. Retrying typically corrects the issue.
This allows `kube_apiserver_insecure_port` to be set to 0 (disabled).
Rework of #1937 with kubeadm support
Also, fixed an issue in `kubeadm-migrate-certs` where the old apiserver cert was copied as the kubeadm key
* Allow setting --bind-address for apiserver hyperkube
This is required if you wish to configure a loadbalancer (e.g haproxy)
running on the master nodes without choosing a different port for the
vip from that used by the API - in this case you need the API to bind to
a specific interface, then haproxy can bind the same port on the VIP:
root@overcloud-controller-0 ~]# netstat -taupen | grep 6443
tcp 0 0 192.168.24.6:6443 0.0.0.0:* LISTEN 0 680613 134504/haproxy
tcp 0 0 192.168.24.16:6443 0.0.0.0:* LISTEN 0 653329 131423/hyperkube
tcp 0 0 192.168.24.16:6443 192.168.24.16:58404 ESTABLISHED 0 652991 131423/hyperkube
tcp 0 0 192.168.24.16:58404 192.168.24.16:6443 ESTABLISHED 0 652986 131423/hyperkube
This can be achieved e.g via:
kube_apiserver_bind_address: 192.168.24.16
* Address code review feedback
* Update kube-apiserver.manifest.j2
* Add Contiv support
Contiv is a network plugin for Kubernetes and Docker. It supports
vlan/vxlan/BGP/Cisco ACI technologies. It support firewall policies,
multiple networks and bridging pods onto physical networks.
* Update contiv version to 1.1.4
Update contiv version to 1.1.4 and added SVC_SUBNET in contiv-config.
* Load openvswitch module to workaround on CentOS7.4
* Set contiv cni version to 0.1.0
Correct contiv CNI version to 0.1.0.
* Use kube_apiserver_endpoint for K8S_API_SERVER
Use kube_apiserver_endpoint as K8S_API_SERVER to make contiv talks
to a available endpoint no matter if there's a loadbalancer or not.
* Make contiv use its own etcd
Before this commit, contiv is using a etcd proxy mode to k8s etcd,
this work fine when the etcd hosts are co-located with contiv etcd
proxy, however the k8s peering certs are only in etcd group, as a
result the etcd-proxy is not able to peering with the k8s etcd on
etcd group, plus the netplugin is always trying to find the etcd
endpoint on localhost, this will cause problem for all netplugins
not runnign on etcd group nodes.
This commit make contiv uses its own etcd, separate from k8s one.
on kube-master nodes (where net-master runs), it will run as leader
mode and on all rest nodes it will run as proxy mode.
* Use cp instead of rsync to copy cni binaries
Since rsync has been removed from hyperkube, this commit changes it
to use cp instead.
* Make contiv-etcd able to run on master nodes
* Add rbac_enabled flag for contiv pods
* Add contiv into CNI network plugin lists
* migrate contiv test to tests/files
Signed-off-by: Cristian Staretu <cristian.staretu@gmail.com>
* Add required rules for contiv netplugin
* Better handling json return of fwdMode
* Make contiv etcd port configurable
* Use default var instead of templating
* roles/download/defaults/main.yml: use contiv 1.1.7
Signed-off-by: Cristian Staretu <cristian.staretu@gmail.com>
* Defaults for apiserver_loadbalancer_domain_name
When loadbalancer_apiserver is defined, use the
apiserver_loadbalancer_domain_name with a given default value.
Fix unconsistencies for checking if apiserver_loadbalancer_domain_name
is defined AND using it with a default value provided at once.
Signed-off-by: Bogdan Dobrelya <bogdando@mail.ru>
* Define defaults for LB modes in common defaults
Adjust the defaults for apiserver_loadbalancer_domain_name and
loadbalancer_apiserver_localhost to come from a single source, which is
kubespray-defaults. Removes some confusion and simplefies the code.
Signed-off-by: Bogdan Dobrelya <bogdando@mail.ru>
Thought this wasn't required at first but I forgot there's no auto flush at the end of these tasks since the `kubernetes/master` role is not the end of the play.
* Fixes an issue where apiserver and friends (controller manager, scheduler) were prevented from restarting after manifests/secrets are changed. This occurred when a replaced kubelet doesn't reconcile new master manifests, which caused old master component versions to linger during deployment. In my case this was causing upgrades from k8s 1.6/1.7 -> k8s 1.8 to fail
* Improves transitions from kubelet container to host kubelet by preventing issues where kubelet container reappeared during the deployment
This allows `kube_apiserver_insecure_port` to be set to 0 (disabled). It's working, but so far I have had to:
1. Make the `uri` module "Wait for apiserver up" checks use `kube_apiserver_port` (HTTPS)
2. Add apiserver client cert/key to the "Wait for apiserver up" checks
3. Update apiserver liveness probe to use HTTPS ports
4. Set `kube_api_anonymous_auth` to true to allow liveness probe to hit apiserver's /healthz over HTTPS (livenessProbes can't use client cert/key unfortunately)
5. RBAC has to be enabled. Anonymous requests are in the `system:unauthenticated` group which is granted access to /healthz by one of RBAC's default ClusterRoleBindings. An equivalent ABAC rule could allow this as well.
Changes 1 and 2 should work for everyone, but 3, 4, and 5 require new coupling of currently independent configuration settings. So I also added a new settings check.
Options:
1. The problem goes away if you have both anonymous-auth and RBAC enabled. This is how kubeadm does it. This may be the best way to go since RBAC is already on by default but anonymous auth is not.
2. Include conditional templates to set a different liveness probe for possible combinations of `kube_apiserver_insecure_port = 0`, RBAC, and `kube_api_anonymous_auth` (won't be possible to cover every case without a guaranteed authorizer for the secure port)
3. Use basic auth headers for the liveness probe (I really don't like this, it adds a new dependency on basic auth which I'd also like to leave independently configurable, and it requires encoded passwords in the apiserver manifest)
Option 1 seems like the clear winner to me, but is there a reason we wouldn't want anonymous-auth on by default? The apiserver binary defaults anonymous-auth to true, but kubespray's default was false.
* Change deprecated vagrant ansible flag 'sudo' to 'become'
* Workaround ansible bug where access var via dict doesn't get real value
When accessing a variable via it's name "{{ foo }}" its value is
retrieved. But when the variable value is retrieved via the vars-dict
"{{ vars['foo'] }}" this doesn't resolve the expression of the variable
any more due to a bug. So e.g. a expression foo="{{ 1 == 1 }}" isn't
longer resolved but just returned as string "1 == 1".
* Make file yamllint complient
Some time ago I think the hardcoded `/var/lib/docker` was required, but kubelet running in a container has been aware of the Docker path since at least as far back as k8s 1.6.
Without this change, you see a large number of errors in the kubelet logs if you installed with a non-default `docker_daemon_graph`
* Refactor downloads to use download role directly
Also disable fact delegation so download delegate works acros OSes.
* clean up bools and ansible_os_family conditionals
* Update main.yml
Needs to set up resolv.conf before updating Yum cache otherwise no name resolution available (resolv.conf empty).
* Update main.yml
Removing trailing spaces
* Add possibility to insert more ip adresses in certificates
* Add newline at end of files
* Move supp ip parameters to k8s-cluster group file
* Add supplementary addresses in kubeadm master role
* Improve openssl indexes
* don't try to install this rpm on fedora atomic
* add docker 1.13.1 for fedora
* built-in docker unit file is sufficient, as tested on both fedora and centos atomic
* Change file used to check kubeadm upgrade method
Test for ca.crt instead of admin.conf because admin.conf
is created during normal deployment.
* more fixes for upgrade
* Rename dns_server to dnsmasq_dns_server so that it includes role prefix
as the var name is generic and conflicts when integrating with existing ansible automation.
* Enable selinux state to be configurable with new var preinstall_selinux_state
PID namespace sharing is disabled only in Kubernetes 1.7.
Explicitily enabling it by default could help reduce unexpected
results when upgrading to or downgrading from 1.7.
This follows pull request #1677, adding the cgroup-driver
autodetection also for kubeadm way of deploying.
Info about this and the possibility to override is added to the docs.
Red Hat family platforms run docker daemon with `--exec-opt
native.cgroupdriver=systemd`. When kubespray tried to start kubelet
service, it failed with:
Error: failed to run Kubelet: failed to create kubelet: misconfiguration: kubelet cgroup driver: "cgroupfs" is different from docker cgroup driver: "systemd"
Setting kubelet's cgroup driver to the correct value for the platform
fixes this issue. The code utilizes autodetection of docker's cgroup
driver, as different RPMs for the same distro may vary in that regard.
New files: /etc/kubernetes/admin.conf
/root/.kube/config
$GITDIR/artifacts/{kubectl,admin.conf}
Optional method to download kubectl and admin.conf if
kubeconfig_lcoalhost is set to true (default false)
* kubeadm support
* move k8s master to a subtask
* disable k8s secrets when using kubeadm
* fix etcd cert serial var
* move simple auth users to master role
* make a kubeadm-specific env file for kubelet
* add non-ha CI job
* change ci boolean vars to json format
* fixup
* Update create-gce.yml
* Update create-gce.yml
* Update create-gce.yml
This sets br_netfilter and net.bridge.bridge-nf-call-iptables sysctl from a single play before kube-proxy is first ran instead of from the flannel and weave network_plugin roles after kube-proxy is started
* using separated vault roles for generate certs with different `O` (Organization) subject field;
* configure vault roles for issuing certificates with different `CN` (Common name) subject field;
* set `CN` and `O` to `kubernetes` and `etcd` certificates;
* vault/defaults vars definition was simplified;
* vault dirs variables defined in kubernetes-defaults foles for using
shared tasks in etcd and kubernetes/secrets roles;
* upgrade vault to 0.8.1;
* generate random vault user password for each role by default;
* fix `serial` file name for vault certs;
* move vault auth request to issue_cert tasks;
* enable `RBAC` in vault CI;
* Added update CA trust step for etcd and kube/secrets roles
* Added load_balancer_domain_name to certificate alt names if defined. Reset CA's in RedHat os.
* Rename kube-cluster-ca.crt to vault-ca.crt, we need separated CA`s for vault, etcd and kube.
* Vault role refactoring, remove optional cert vault auth because not not used and worked. Create separate CA`s fro vault and etcd.
* Fixed different certificates set for vault cert_managment
* Update doc/vault.md
* Fixed condition create vault CA, wrong group
* Fixed missing etcd_cert_path mount for rkt deployment type. Distribute vault roles for all vault hosts
* Removed wrong when condition in create etcd role vault tasks.
* Updates Controller Manager/Kubelet with Flannel's required configuration for CNI
* Removes old Flannel installation
* Install CNI enabled Flannel DaemonSet/ConfigMap/CNI bins and config (with portmap plugin) on host
* Uses RBAC if enabled
* Fixed an issue that could occur if br_netfilter is not a module and net.bridge.bridge-nf-call-iptables sysctl was not set
* Adding yaml linter to ci check
* Minor linting fixes from yamllint
* Changing CI to install python pkgs from requirements.txt
- adding in a secondary requirements.txt for tests
- moving yamllint to tests requirements
If Kubernetes > 1.6 register standalone master nodes w/ a
node-role.kubernetes.io/master=:NoSchedule taint to allow
for more flexible scheduling rather than just marking unschedulable.
Change kubelet deploy mode to host
Enable cri and qos per cgroup for kubelet
Update CoreOS images
Add upgrade hook for switching from kubelet deployment from docker to host.
Bump machine type for ubuntu-rkt-sep
* Added custom ips to etcd vault distributed certificates
* Added custom ips to kube-master vault distributed certificates
* Added comment about issue_cert_copy_ca var in vault/issue_cert role file
* Generate kube-proxy, controller-manager and scheduler certificates by vault
* Revert "Disable vault from CI (#1546)"
This reverts commit 781f31d2b8.
* Fixed upgrade cluster with vault cert manager
* Remove vault dir in reset playbook
* Bump tag for upgrade CI, fix netchecker upgrade
netchecker-server was changed from pod to deployment, so
we need an upgrade hook for it.
CI now uses v2.1.1 as a basis for upgrade.
* Fix upgrades for certs from non-rbac to rbac
When VPC is used, external DNS might not be available. This patch change
behavior to use metadata service instead of external DNS when
upstream_dns_servers is not specified.
Signed-off-by: Sergii Golovatiuk <sgolovatiuk@mirantis.com>
According to code apiserver, scheduler, controller-manager, proxy don't
use resolution of objects they created. It's not harmful to change
policy to have external resolver.
Signed-off-by: Sergii Golovatiuk <sgolovatiuk@mirantis.com>
In kubernetes 1.6 ClusterFirstWithHostNet was added as an option. In
accordance to it kubelet will generate resolv.conf based on own
resolv.conf. However, this doesn't create 'options', thus the proper
solution requires some investigation.
This patch sets the same resolv.conf for kubelet as host
Signed-off-by: Sergii Golovatiuk <sgolovatiuk@mirantis.com>
- Run docker run from script rather than directly from systemd target
- Refactoring styling/templates
Signed-off-by: Sergii Golovatiuk <sgolovatiuk@mirantis.com>
Non-brekable space is 0xc2 0xa0 byte sequence in UTF-8.
To find one:
$ git grep -I -P '\xc2\xa0'
To replace with regular space:
$ git grep -l -I -P '\xc2\xa0' | xargs sed -i 's/\xc2\xa0/ /g'
This commit doesn't include changes that will overlap with commit f1c59a91a1.
By default Calico CNI does not create any network access policies
or profiles if 'policy' is enabled in CNI config. And without any
policies/profiles network access to/from PODs is blocked.
K8s related policies are created by calico-policy-controller in
such case. So we need to start it as soon as possible, before any
real workloads.
This patch also fixes kube-api port in calico-policy-controller
yaml template.
Closes#1132
It is now possible to deactivate selected authentication methods
(basic auth, token auth) inside the cluster by adding
removing the required arguments to the Kube API Server and generating
the secrets accordingly.
The x509 authentification is currently not optional because disabling it
would affect the kubectl clients deployed on the master nodes.
Default backend is now etcd3 (was etcd2).
The migration process consists of the following steps:
* check if migration is necessary
* stop etcd on first etcd server
* run migration script
* start etcd on first etcd server
* stop kube-apiserver until configuration is updated
* update kube-apiserver
* purge old etcdv2 data
Issue #1125. Make RBAC authorization plugin work out of the box.
"When bootstrapping, superuser credentials should include the system:masters group, for example by creating a client cert with /O=system:masters. This gives those credentials full access to the API and allows an admin to then set up bindings for other users."
To use OpenID Connect Authentication beside deploying an OpenID Connect
Identity Provider it is necesarry to pass additional arguments to the Kube API Server.
These required arguments were added to the kube apiserver manifest.
Until now it was not possible to add an API Loadbalancer
without an static IP Address. But certain Loadbalancers
like AWS Elastic Loadbalanacer dontt have an fixed IP address.
With this commit it is possible to add these kind of Loadbalancers
to the Kargo deployment.
Updates based on feedback
Simplify checks for file exists
remove invalid char
Review feedback. Use regular systemd file.
Add template for docker systemd atomic
Kubernetes project is about to set etcdv3 as default storage engine in
1.6. This patch allows to specify particular backend for
kube-apiserver. User may force the option to etcdv3 for new environment.
At the same time if the environment uses v2 it will continue uses it
until user decides to upgrade to v3.
Signed-off-by: Sergii Golovatiuk <sgolovatiuk@mirantis.com>
* Leave all.yml to keep only optional vars
* Store groups' specific vars by existing group names
* Fix optional vars casted as mandatory (add default())
* Fix missing defaults for an optional IP var
* Relink group_vars for terraform to reflect changes
Signed-off-by: Bogdan Dobrelya <bogdando@mail.ru>
Sometimes, a sysadmin might outright delete the SELinux rpms and
delete the configuration. This causes the selinux module to fail
with
```
IOError: [Errno 2] No such file or directory: '/etc/selinux/config'\n",
"module_stdout": "", "msg": "MODULE FAILURE"}
```
This simply checks that /etc/selinux/config exists before we try
to set it Permissive.
Update from feedback
Migrate older inline= syntax to pure yml syntax for module args as to be consistant with most of the rest of the tasks
Cleanup some spacing in various files
Rename some files named yaml to yml for consistancy
Kubelet is responsible for creating symlinks from /var/lib/docker to /var/log
to make fluentd logging collector work.
However without using host's /var/log those links are invisible to fluentd.
This is done on rkt configuration too.
"shell" step doesn't support check mode, which currently leads to failures,
when Ansible is being run in check mode (because Ansible doesn't run command,
assuming that command might have effect, and no "rc" or "output" is registered).
Setting "check_mode: no" allows to run those "shell" commands in check mode
(which is safe, because those shell commands doesn't have side effects).
always_run was deprecated in Ansible 2.2 and will be removed in 2.4
ansible logs contain "[DEPRECATION WARNING]: always_run is deprecated.
Use check_mode = no instead". This patch fix deprecation.
Since systemd kubelet.service has {{ ssl_ca_dirs }}, fact should be
gathered before writing kubelet.service.
Closes: #1007
Signed-off-by: Sergii Golovatiuk <sgolovatiuk@mirantis.com>
- Exclude kubelet CPU/RAM (kube-reserved) from cgroup. It decreases a
chance of overcommitment
- Add a possibility to modify Kubelet node-status-update-frequency
- Add a posibility to configure node-monitor-grace-period,
node-monitor-period, pod-eviction-timeout for Kubernetes controller
manager
- Add Kubernetes Relaibility Documentation with recomendations for
various scenarios.
Signed-off-by: Sergii Golovatiuk <sgolovatiuk@mirantis.com>
kubelet lost the ability to load kernel modules. This
puts that back by adding the lib/modules mount to kubelet.
The new variable kubelet_load_modules can be set to true
to enable this item. It is OFF by default.
Netchecker is rewritten in Go lang with some new args instead of
env variables. Also netchecker-server no longer requires kubectl
container. Updating playbooks accordingly.
- Docker 1.12 and further don't need nsenter hack. This patch removes
it. Also, it bumps the minimal version to 1.12.
Closes#776
Signed-off-by: Sergii Golovatiuk <sgolovatiuk@mirantis.com>
* Drop linux capabilities for unprivileged containerized
worlkoads Kargo configures for deployments.
* Configure required securityContext/user/group/groups for kube
components' static manifests, etcd, calico-rr and k8s apps,
like dnsmasq daemonset.
* Rework cloud-init (etcd) users creation for CoreOS.
* Fix nologin paths, adjust defaults for addusers role and ensure
supplementary groups membership added for users.
* Add netplug user for network plugins (yet unused by privileged
networking containers though).
* Grant the kube and netplug users read access for etcd certs via
the etcd certs group.
* Grant group read access to kube certs via the kube cert group.
* Remove priveleged mode for calico-rr and run it under its uid/gid
and supplementary etcd_cert group.
* Adjust docs.
* Align cpu/memory limits and dropped caps with added rkt support
for control plane.
Signed-off-by: Bogdan Dobrelya <bogdando@mail.ru>
be run by limit on each node without regard for order.
The changes make sure that all of the directories needed to do
certificate management are on the master[0] or etcd[0] node regardless
of when the playbook gets run on each node. This allows for separate
ansible playbook runs in parallel that don't have to be synchronized.
the openssl tools will fail to create signing requests because
the CN is too long. This is mainly a problem when FQDNs are used
in the inventory file.
THis will truncate the hostname for the CN field only at the
first dot. This should handle the issue for most cases.
Also remove the check for != "RedHat" when removing the dhclient hook,
as this had also to be done on other distros. Instead, check if the
dhclienthookfile is defined.
the tasks fail because selinux prevents ip-forwarding setting.
Moving the tasks around addresses two issues. Makes sure that
the correct python tools are in place before adjusting of selinux
and makes sure that ipforwarding is toggled after selinux adjustments.
* Add restart for weave service unit
* Reuse docker_bin_dir everythere
* Limit systemd managed docker containers by CPU/RAM. Do not configure native
systemd limits due to the lack of consensus in the kernel community
requires out-of-tree kernel patches.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Current design expects users to define at least one
nameserver in the nameservers var to backup host OS DNS config
when the K8s cluster DNS service IP is not available and hosts
still have to resolve external or intranet FQDNs.
Fix undefined nameservers to fallback to the default_resolver.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Add BGP route reflectors support in order to optimize BGP topology
for deployments with Calico network plugin.
Also bump version of calico/ctl for some bug fixes.
Also place in global vars and do not repeat the kube_*_config_dir
and kube_namespace vars for better code maintainability and UX.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Do not repeat options and nameservers in the dhclient hooks.
Do not prepend nameservers for dhclient but supersede and fail back
to the upstream_dns_resolvers then default_resolver. Fixes order of
nameservers placement, which is cluster DNS ip goes always first.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
* For Debian/RedHat OS families (with NetworkManager/dhclient/resolvconf
optionally enabled) prepend /etc/resolv.conf with required nameservers,
options, and supersede domain and search domains via the dhclient/resolvconf
hooks.
* Drop (z)nodnsupdate dhclient hook and re-implement it to complement the
resolvconf -u command, which is distro/cloud provider specific.
Update docs as well.
* Enable network restart to apply and persist changes and simplify handlers
to rely on network restart only. This fixes DNS resolve for hostnet K8s
pods for Red Hat OS family. Skip network restart for canal/calico plugins,
unless https://github.com/projectcalico/felix/issues/1185 fixed.
* Replace linefiles line plus with_items to block mode as it's faster.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Co-authored-by: Matthew Mosesohn <mmosesohn@mirantis.com>
In order to enable offline/intranet installation cases:
* Move DNS/resolvconf configuration to preinstall role. Remove
skip_dnsmasq_k8s var as not needed anymore.
* Preconfigure DNS stack early, which may be the case when downloading
artifacts from intranet repositories. Do not configure
K8s DNS resolvers for hosts /etc/resolv.conf yet early (as they may be
not existing).
* Reconfigure K8s DNS resolvers for hosts only after kubedns/dnsmasq
was set up and before K8s apps to be created.
* Move docker install task to early stage as well and unbind it from the
etcd role's specific install path. Fix external flannel dependency on
docker role handlers. Also fix the docker restart handlers' steps
ordering to match the expected sequence (the socket then the service).
* Add default resolver fact, which is
the cloud provider specific and remove hardcoded GCE resolver.
* Reduce default ndots for hosts /etc/resolv.conf to 2. Multiple search
domains combined with high ndots values lead to poor performance of
DNS stack and make ansible workers to fail very often with the
"Timeout (12s) waiting for privilege escalation prompt:" error.
* Update docs.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
The variale etcd_access_addresses is used to determine
how to address communication from other roles to
the etcd cluster.
It was set to the address that ansible uses to
connect to instance ({{ item }})s and not the
the variable:
ip_access
which had already been created and could already
be overridden through the access_ip variable.
This change allows ansible to connect to a machine using
a different address than the one used to access etcd.
Override GCE sysctl in /etc/sysctl.d/99-sysctl.conf instead of
the /etc/sysctl.d/11-gce-network-security.conf. It is recreated
by GCE, f.e. if gcloud CLI invokes some security related changes,
thus losing customizations we want to be persistent.
Update cloud providers firewall requirements in calico docs.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Fixes#655.
This is a teporary solution for long-polling idle connections to
apiserver. It will make Nginx not cut them for the duration of expected
timeout. It will also make Nginx extremely slow in realizing that there
is some issue with connectivity to apiserver as well, so it might not be
perfect permanent solution.
* Add an option to deploy K8s app to test e2e network connectivity
and cluster DNS resolve via Kubedns for nethost/simple pods
(defaults to false).
* Parametrize existing k8s apps templates with kube_namespace and
kube_config_dir instead of hardcode.
* For CoreOS, ensure nameservers from inventory to be put in the
first place to allow hostnet pods connectivity via short names
or FQDN and hostnet agents to pass as well, if netchecker
deployed.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
* Add dns_replicas, dns_memory/cpu_limit/requests vars for
dns related apps.
* When kube_log_level=4, log dnsmasq queries as well.
* Add log level control for skydns (part of kubedns app).
* Add limits/requests vars for dnsmasq (part of kubedns app) and
dnsmasq daemon set.
* Drop string defaults for kube_log_level as it is int and
is defined in the global vars as well.
* Add docs
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
W/o this patch, the "Download containers" task may be skipped
when running on the delegate node due to wrong "when" confition.
Then it fails to upload nginx image to the nodes as well.
Fix download nginx dependency so it always can be pushed to
nodes when download_run_once is enabled.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
According to http://kubernetes.io/docs/user-guide/images/ :
By default, the kubelet will try to pull each image from the
specified registry. However, if the imagePullPolicy property
of the container is set to IfNotPresent or Never, then a local\
image is used (preferentially or exclusively, respectively).
Use IfNotPresent value to allow images prepared by the download
role dependencies to be effectively used by kubelet without pull
errors resulting apps to stay blocked in PullBackOff/Error state
even when there are images on the localhost exist.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Pre download all required container images as roles' deps.
Drop unused flannel-server-helper images pre download.
Improve pods creation post-install test pre downloaded busybox.
Improve logs collection script with kubectl describe, fix sudo/etcd/weave
commands.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Fix unreliable waiting for the apiserver to become ready.
Remove logfile mount to align with the rest of static pods
and because containers shall write logs to stdout only.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
- Move CNI configuration from `kubernetes/node` role to
`network_plugin/canal`
- Create SSL dir for Canal and symlink etcd SSL files
- Add needed options to `canal-config` configmap
- Run flannel and calico-node containers with proper configuration
- Drop debugs from collect-info playbook
- Drop sudo from collect-info step and add target dir var (required for travis jobs)
- Label all k8s apps, including static manifests
- Add logs for K8s apps to be collected as well
- Fix upload to GCS as a public-read tarball
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Related bug: https://github.com/ansible/ansible/issues/15405
Uses tar and register because synchronize module cannot sudo on the
remote side correctly and copy is too slow.
This patch dramatically cuts down the number of tasks to process
for cert synchronization.
* Don't push containers if not changed
* Do preinstall role only once and redistribute defaults to
corresponding roles
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Change the kubelet --hostname-override flag to use the ansible_hostname variable which should be more consistent with the value required by cloud providers
Add ansible_hostname alias to /etc/hosts when it is different from inventory_hostname to overcome node name limitations see https://github.com/kubernetes/kubernetes/issues/22770
Signed-off-by: Chad Swenson <chadswen@gmail.com>
The requirements for network policy feature are described here [1]. In
order to enable it, appropriate configuration must be provided to the CNI
plug in and Calico policy controller must be set up. Beside that
corresponding extensions needed to be enabled in k8s API.
Now to turn on the feature user can define `enable_network_policy`
customization variable for Ansible.
[1] http://kubernetes.io/docs/user-guide/networkpolicies/
Also adds all masters by hostname and localhost/127.0.0.1 to
apiserver SSL certificate.
Includes documentation update on how localhost loadbalancer works.
* Add a var for ndots (default 5) and put it hosts' /etc/resolv.conf.
* Poke kube dns container image to v1.7
* In order to apply changes to kubelet, notify it to
be restarted on changes made to /etc/resolv.conf. Ignore errors as the kubelet
may yet to be present up to the moment of the notification being processed.
* Remove unnecessary kubelet restart for master role as the node role ensures
it is up and running. Notify master static pods waiters for apiserver,
scheduler, controller-manager instead.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Change additional dnsmasq opts:
- Adjust caching size and TTL
- Disable resolve conf to not create loops
- Change dnsPolicy to default (similarly to kubedns's dnsmasq). The
ClusterFirst should not be used to not create loops
- Disable negative NXDOMAIN replies to be cached
- Make its very installation as optional step (enabled by default).
If you don't want more than 3 DNS servers, including 1 for K8s, disable
it.
- Add docs and a drawing to clarify DNS setup.
- Fix stdout logs for dnsmasq/kubedns app configs
- Add missed notifies to resolvconf -u handler
- Fix idempotency of resolvconf head file changes
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
* Add the retry_stagger var to tweak push and retry time strategies.
* Add large deployments related docs.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Move version/repo vars to download role.
Add container to download params, which overrides url/source_url,
if enabled.
Fix networking plugins download depending on kube_network_plugin.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Shorten deployment time with:
- Remove redundand roles if duplicated by a dependency and vice versa
- When a member of k8s-cluster, always install docker as a dependency
of the etcd role and drop the docker role from cluster.yaml.
- Drop etcd and node role dependencies from master role as they are
covered by the node role in k8s-cluster group as well. Copy defaults
for master from node role.
- Decouple master, node, secrets roles handlers and vars to be used w/o
cross references.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Creating the unit using default settings early on
and then changing it during network_plugin section
leads to too many docker restarts and duplicated code.
Reversed Wants= dependence on docker.service so it does not
restart docker when reloading systemd
Consolidated all docker restart handlers.
* Add for docker system units:
ExecReload=/bin/kill -s HUP $MAINPID
Delegate=yes
KillMode=process.
* Add missed DOCKER_OPTIONS for calico/weave docker systemd unit.
* Change Requires= to a less strict and non-faily Wants=, add missing
Wants= for After=.
* Align wants/after in a wat if Wants=foo, After= has foo as well.
* Make wants/after docker.service to ask for the docker.socket as well.
* Move "docker rm -f" commands from ExecStartPre= to ExecStopPost=.
hooks to ensure non-destructive start attempts issued by Wants=.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
etcd facts are generated in kubernetes/preinstall, so etcd nodes need
to be evaluated first before the rest of the deployment.
Moved several directory facts from kubernetes/node to
kubernetes/preinstall because they are not backward dependent.
* Add HA docs for API server.
* Add auto-evaluated internal endpoints and clarify the loadbalancer_apiserver
vars and usecases.
* Use facts for kube_apiserver to not repeat code and enable LB endpoints use.
* Use /healthz check for the wait-for apiserver.
* Use the single endpoint for kubelet instead of the list of apiservers
* Specify kube_apiserver_count to for HA layout
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Nearly the last stage of source all components to containers.
Kubectl will be called from hyperkube image.
Remaining tasks:
* Move kube_version variable to kubernetes/preinstall
* Drop placeholder download.nothing requirement
kubelet via docker
kube-apiserver as a static pod
Fixed etcd service start to be more tolerant of slow start.
Workaround for kube_version to stay in download role, but not
download an files by creating a new "nothing" download entry.
* Add auto-evaluated internal endpoints and clarify the loadbalancer_apiserver
vars and usecases.
* Add loadbalancer_apiserver_localhost (default false). If enabled, override
the external LB and expect localhost:443/8080 to be new internal only frontends.
* Add kube_apiserver_multiaccess to ignore loadbalancers, and make clients
to access the apiservers as a comma-separated list of access_ip/ip/ansible ip
(a default mode). When disabled, allow clients to use the given loadbalancers.
* Define connections security mode for kube controllers, schedulers, proxies.
It is insecure be default, which is the current deployment choice.
* Rework the groups['kube-master'][0] hardcode defining the apiserver
endpoints.
* Improve grouping of vars and add facts for kube_apiserver.
* Define kube_apiserver_insecure_bind_address as a fact, add more
facts for ease of use.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Move set_facts to the preinstall scope, so every role
may see it. For example, network plugins to see the etcd_endpoint.
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
* Enforce a etcd-proxy role to a k8s-cluster group members. This
provides an HA layout for all of the k8s cluster internal clients.
* Proxies to be run on each node in the group as a separate etcd
instances with a readwrite proxy mode and listen the given endpoint,
which is either the access_ip:2379 or the localhost:2379.
* A notion for the 'kube_etcd_multiaccess' is: ignore endpoints and
loadbalancers and use the etcd members IPs as a comma-separated
list. Otherwise, clients shall use the local endpoint provided by a
etcd-proxy instances on each etcd node. A Netwroking plugins always
use that access mode.
* Fix apiserver's etcd servers args to use the etcd_access_endpoint.
* Fix networking plugins flannel/calico to use the etcd_endpoint.
* Fix name env var for non masters to be set as well.
* Fix etcd_client_url was not used anywhere and other etcd_* facts
evaluation was duplicated in a few places.
* Define proxy modes only in the env file, if not a master. Del
an automatic proxy mode decisions for etcd nodes in init/unit scripts.
* Use Wants= instead of Requires= as "This is the recommended way to
hook start-up of one unit to the start-up of another unit"
* Make apiserver/calico Wants= etcd-proxy to keep it always up
Signed-off-by: Bogdan Dobrelya <bdobrelia@mirantis.com>
Co-authored-by: Matthew Mosesohn <mmosesohn@mirantis.com>
