When etcd exceeds its memory limit, it becomes useless but keeps running.
We should let OOM killer kill etcd process in the container, so systemd can spot
the problem and restart etcd according to "Restart" setting in etcd.service unit file.
If OOME problem keep repeating, i.e. it happens every single restart,
systemd will eventually back off and stop restarting it anyway.
--restart=on-failure:5 in this file has no effect because memory allocation error
doesn't by itself cause the process to die
Related: https://github.com/kubernetes-incubator/kubespray/blob/master/roles/etcd/templates/etcd-docker.service.j2
This kind of reverts a change introduced in #1860.
- Run docker run from script rather than directly from systemd target
- Refactoring styling/templates
Signed-off-by: Sergii Golovatiuk <sgolovatiuk@mirantis.com>
Reduce election timeout to 5000ms (was 10000ms)
Raise heartbeat interval to 250ms (was 100ms)
Remove etcd cpu share (was 300)
Make etcd_cpu_limit and etcd_memory_limit optional.
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.
* 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>
Running etcd in Docker reduces the number of individual file
downloads and services running on the host.
Note: etcd container v3.0.1 moves bindir to /usr/local/bin
Fixes: #298
* Set ETCD_INITIAL_CLUSTER_STATE from `new` to `existing`,
because parameter `new` makes sense only on cluster assembly
stage.
* If cluster exists and current node is not a part
of the cluster, add it with command `etcdctl add member name url`.
Closes kubespray/kargo/#270
Each node can have 3 IPs.
1. ansible_default_ip4 - whatever ansible things is the first IPv4 address
usually with the default gw.
2. ip - An address to use on the local node to bind listeners and do local
communication. For example, Vagrant boxes have a first address that is the
NAT bridge and is common for all nodes. The second address/interface should
be used.
3. access_ip - An address to use for node-to-node access. This is assumed to
be used by other nodes to access the node and may not be actually assigned
on the node. For example, AWS public ip that is not assigned to node.
This updates the places addresses are used to use either ip or access_ip and walk
up the list to find an address.
fix etcd configuration for nodes
fix wrong calico checksums
using a var name etcd_bin_dir
fix etcd handlers for sysvinit
using a var name etcd_bin_dir
sysvinit script
review etcd configuration
