Snap refresh
Charms that comprise Charmed Kubernetes install required software such
as etcd
, kubelet
, and kube-apiserver
using snap packages. The snapd
daemon periodically scans installed snaps for updates and will automatically
refresh upgradeable packages to ensure the software is current.
All charms built from layer-snap include a snapd_refresh
configuration
option that can be used to adjust the snapd
refresh interval. By default,
Charmed Kubernetes charms set this option to the maximum amount of time
that snapd
will allow between scans.
Refresh interval configuration
Display the currently configured snapd_refresh
option for a given charm:
juju config <charm> snapd_refresh
Change the snapd_refresh
option with the following:
juju config <charm> snapd_refresh=<value>
The value returned or set above should be an explicit timer, an empty string, or the special keyword max.
Explicit timer
An explicit timer may be a simple
mon
(scan every Monday) or a more complexmon3,23:00
(scan on the third Monday of the month at 23:00). See possible values for explicit timers in the refresh.timer section of the system options documentation.Empty string
An empty string instructs
snapd
to refresh snap packages according to the default system policy. This is currently 4 times per day.max
When set to
max
, refresh scans will be delayed for the maximum amount of time allowed bysnapd
. This is currently once per month based on the date this option was set.
Determine the actual max value
Use snap get
on a deployed system to determine the value that snapd
uses
when a charm is configured with snapd_refresh=max
. An example with etcd
shows that the max
option used when etcd/0
was deployed has mapped to the
last Sunday of every month (sun5
):
juju config etcd snapd_refresh
max
juju exec --unit etcd/0 'snap get system refresh.timer'
sun5
Refresh interval among peers
Charmed Kubernetes applications that support peering will use Juju
leadership to configure a consistent refresh interval among peers. The lead
unit for etcd
, kubernetes-control-plane
, and kubernetes-worker
applications
will set an initial refresh value. Subsequent units that join as followers
will use the leader value as their snap refresh interval. This ensures all
units in a peer group will refresh at approximately the same time.
Force snapd to refresh installed snaps
If an immediate snap refresh is desired, invoke snap refresh
on the
applicable cluster components. For example, refresh all snaps on all etcd
units with:
juju exec --application etcd 'snap refresh'
As another example, only refresh the cdk-addons
snap on the
kubernetes-control-plane/0
unit with:
juju exec --unit kubernetes-control-plane/0 'snap refresh cdk-addons'
See the guide to contributing or discuss these docs in our public Mattermost channel.