Previously, this chart was not deployable in a Kubernetes namespace with
[Resource Quotas](https://kubernetes.io/docs/concepts/policy/resource-quotas/)
enabled. A user could set resource limits on the CRDB pods, but not on the
short-lived pod that is part of the cockroachdb-init job. This adds an
additional resource configuration to that pod spec to allow this chart to be
deployed when Resource Quotas are enabled.
Signed-off-by: Joel Kenny <joel@cockroachlabs.com>
Updated README to direct to CockroachDB documentation on the Helm chart, and to provide updated upgrade guidance.
Checklist
[Place an '[x]' (no spaces) in all applicable fields. Please remove unrelated fields.]
[x] DCO signed
[x] Variables are documented in the README.md
[x] Title of the PR starts with chart name (e.g. [stable/chart])
Signed-off-by: Keith McClellan <keith.mcclellan@gmail.com>
modified: Chart.yaml
Signed-off-by: Keith McClellan <keith.mcclellan@gmail.com>
These three all have good knowledge of CockroachDB and experience
running it in production on Kubernetes.
Signed-off-by: Alex Robinson <alexdwanerobinson@gmail.com>
Sets client label on init job when NetworkPolicy.Enabled and !NetworkPolicy.AllowExternal
Signed-off-by: Samuel Chantaraud <samuel.chantaraud@messagepoint.com>
In the past there had problems with Minikube not accepting large disk sizes,
but it no longer seems to be a problem on more recent versions, so let's
default to something a little more reasonable here.
Signed-off-by: Bram Gruneir <bram@cockroachlabs.com>
The default disk size of 1 GiB can cause mistakes (e.g.
https://stackoverflow.com/q/52245811/1925481), but is needed in
order to work in environments like Minikube.
And I don't know what's going on with the way that line breaks work in
NOTES.txt, but this is what makes it render properly.
Signed-off-by: Alex Robinson <alexdwanerobinson@gmail.com>
* Add missing NodeSelector option
* Replace obsolete Cpu and Memory options with Resources option
* Touch up a couple descriptions and defaults, reorder to better match
values in values.yaml
Signed-off-by: Alex Robinson <alexdwanerobinson@gmail.com>
This was preventing any upgrades because kubernetes made the selector
immutable and the chart version changes between upgrades of charts.
fix#8102
Signed-off-by: Rio Kierkels <riokierkels@gmail.com>
* Added functionality to let us join an already existing cluster.
Signed-off-by: Oliver Isaac <oisaac@gmail.com>
* Added locality to values.yaml and README. Version bump.
Signed-off-by: Oliver Isaac <oisaac@gmail.com>
* Cleaned up trailing whitespace
Signed-off-by: Oliver Isaac <oisaac@gmail.com>
* Typo and join bug-fix
Signed-off-by: Oliver Isaac <oisaac@gmail.com>
* [stable/cockroachdb] Added support for node selector
* Merge with charts/stable
* [stable/cockroachdb] Added support for node selector
* [stable/cockroachdb] Added support for node selector
new chart stable/mysqldump (#4069)
* new chart stable/mysqldump
Helps the user backup mysql databases to a persistent volume
* use chart.name from tpl, add app label
* add app/chart fixes to mysqldump-cron.yaml
* fix backoffLimit in wrong place.
* exit gracefully if db host not specified
Merge with charts/stable
* Chart version bump
* Added external and internal GRPC port as options. Also added options to customize service annotations. Need these for istio config.
* Updated README to include values I added. Reformatted the table so it reads pretty in plaintext
* cockroach chart version bump for istio work
* [stable/cockroachdb] Added support for node selector
* Merge with charts/stable
* [stable/cockroachdb] Added support for node selector
* [stable/cockroachdb] Added support for node selector
new chart stable/mysqldump (#4069)
* new chart stable/mysqldump
Helps the user backup mysql databases to a persistent volume
* use chart.name from tpl, add app label
* add app/chart fixes to mysqldump-cron.yaml
* fix backoffLimit in wrong place.
* exit gracefully if db host not specified
Merge with charts/stable
* Chart version bump
We had a deadlock of sorts. --wait waits until all pods are ready.
postinstall hooks don't run until after --wait is done waiting. The pods
won't be ready until the `init` job runs, which was a postinstall hook.
This gets around that by making the init command run as part of the
chart rather than as a postinstall hook. I added a bash retry loop for
the init command to get around the lack of ability to configure pod
restart backoffs in Kubernetes.
The containers in the stateful set were ignoring the resource requests
when specified in `values.yaml`. Comment out the resources requests
in `values.yaml`, so we only set requests if the user specifically
enables them.
I verified this is working by running `kubectl describe pod POD | grep
cpu` on one of the created pods on the fix branch. It returned `cpu:
"100m"`. On the master branch, the same command returned nothing.
* cockroachdb: Add readiness/liveness probes
* cockroachdb: Switch from unready-endpoints annotation to proper field
This works as far back as Kubernetes 1.8, which means we're still
supporting the 3 most recently release versions of Kubernetes, more than
the 2 required by the README of this repo.
However, I'm leaving the annotation in because the field itself doesn't
work on 1.9.x: https://github.com/kubernetes/kubernetes/issues/58662
* cockroachdb: Add podManagementPolicy and UpdateStrategy
This is important when testing against alpha and beta builds of
Kubernetes along with environments that use pre-releases to denote
things other than pre-releases (e.g., gke denotes the environment
with a pre-releases)
* Fixing version comparisons from strings to semver
This impacts the jump to kubernetes 1.10. For more information
see issue 3002
* Updating chart owner to github id
* Bump CockroachDB chart to v1.1.4
* Switch to safer initialization mechanism in CockroachDB chart
This removes the dependency on the peer-finder-based init container,
replacing it with a post-install hook that runs the `cockroach init`
command for initializing a CockroachDB cluster.
This removes a (very rare) potential edge case where a cluster could
end up in split brain if all pods went down, cockroachdb-0 lost its
disk, and the other pods didn't.
* cockroachdb: Change from alpha annotation to proper StorageClass field
* cockroachdb: Add test to verify cluster is reachable and functional
* Add options for configuring cache and SQL memory to CockroachDB chart
* Bump CockroachDB version to v1.1.3
* Improve CockroachDB's PodDisruptionBudget
Letting only 1 go down is much safer and more accurately reflects the
desire of operators than allowing up to 33% to go down. It's
unfortunate that this bumps the minimum version by so much, but 1.7 is a
full two releases ago.
Rebecca Taft
Joel Kenny
Andrei Matei
keith-mcclellan
keith-mcclellan
Alex Robinson
Carlos Tadeu Panato Junior
Christian Hüning
Solon
Jonas Eckerström
Nik Voss
RaduBerinde
Bob Vawter
Ben Darnell
schantaraud
Nathan VanBenschoten
Bram Gruneir
Rio Kierkels
Nikhil Benesch
Oliver
Brandon Harper
Matt McNaughton
Matt Farina
AdamDang
Matt Cholick
Lachlan Evenson
Roy Hvaara
Joey Pereira
kevinschumacher
Jordan Lewis
Sameer Naik
Dan (Turk)