* added notification when uncordoning
when reboot & uncordoning is succ
essful -> notification will be se
nt
* added uncordon message tmpl
added message template for
announcing successful uncor-
doning and reboot.
* added proper documentation about new flag
added readme note about new flag
* Added support for multi-arch image build
* Requested changes to multi-arch build
* Further optimizations of multi build
* multi needs QEMU for some pieces
* change main push for all platforms
* Update Dockerfile to call Makefile
* Remove manual workflow
Terminated pods should be excluded from the blocking a reboot as per https://github.com/weaveworks/kured/issues/227
This adds status filters to the fieldSelector in order to do that. I've not updated tests here but have successfully tested the exact same filter using kubectl
This commit introduces a new flag '--log-format' that allows a user
to configure json logging on the pods. If the log-format
is not specified, the formatter will default to the existing
text formatter.
In this PR the slack-hook-url is translated
into shoutrrr syntax. Therefore, slack pack
age as well as checks for slack-hook-url in
drain and reboot functions are removed.
Also added a unit test for flagCheck(), this
function also checks the (slack)URL syntax.
Currently, kured issues the system reboot command immediately after
kubectl drain finishes.
This is a problem for processes that need extra time to finish but aren't
running on pods and therefore aren't controlled by kubectl drain (e.g.
de-registering nodes from external load balancers).
This patch solves the problem by introducing a `reboot-delay` command
line argument that can be used to add a delay after kubectl drain
finishes but before the reboot command is issued.
* prometheus labels incl tests
* enable label in main, add log, docs
* revert the option to query by label
* revert the option to query by label
* PromClient instantiate by func,white space removal
* revert whitespace fix for readability.
* revert removal of newlines for readability
* rename New to NewPromClient to improve readability
Co-authored-by: simp <simp@saxobank.com>
This support throtteling of reboots across the cluster
and allows rebooted nodes to reschedule pods, e.g.
to synchronize replicated state before rebooting the next node.
Without this patch, the rebootCommand passed to invokeReboot is
ignored, and the command used for reboot is always systemctl reboot.
This is a problem, as we are aiming for flexible commands for this
release.
This fixes it by restoring the previous behaviour before commit
[1] happened.
[1]: 694957d56e
