The following commands reference manifests that come from the same repository as the source code is being built from. This means we can't have a pinned reference to the latest release as that is a chicken-egg problem. Therefore, we use a rolling tag for the particular branch in our manifests. :::{caution} For production deployment, you should always replace the {{productName}} image in all the manifests that contain it with a stable (full version) reference. We'd encourage you to use a sha reference, although using full-version tags is also fine. :::
{{productName}} has a few dependencies that you need to install to your cluster first.
In case you already have a supported version of each of these dependencies installed in your cluster, you can skip this part.
:::{code-block} shell :substitutions:
kubectl apply --server-side -f=https://raw.githubusercontent.com/{{repository}}/{{revision}}/examples/third-party/cert-manager.yaml :::
:::{code-block} shell
kubectl wait --for condition=established --timeout=60s crd/certificates.cert-manager.io crd/issuers.cert-manager.io
for deploy in cert-manager{,-cainjector,-webhook}; do kubectl -n=cert-manager rollout status --timeout=10m deployment.apps/"${deploy}" done
for i in {1..30}; do { kubectl -n=cert-manager get secret/cert-manager-webhook-ca && break; } || sleep 1 done :::
:::{code-block} shell :substitutions:
kubectl -n=prometheus-operator apply --server-side -f=https://raw.githubusercontent.com/{{repository}}/{{revision}}/examples/third-party/prometheus-operator.yaml :::
:::{code-block} shell
kubectl wait --for='condition=established' crd/prometheuses.monitoring.coreos.com crd/prometheusrules.monitoring.coreos.com crd/servicemonitors.monitoring.coreos.com
kubectl -n=prometheus-operator rollout status --timeout=10m deployment.apps/prometheus-operator
for i in {1..30}; do { kubectl -n=cert-manager get secret/cert-manager-webhook-ca && break; } || sleep 1 done :::
Once you have the dependencies installed and available in your cluster, it is the time to install {{productName}}.
:::{code-block} shell :substitutions:
kubectl -n=scylla-operator apply --server-side -f=https://raw.githubusercontent.com/{{repository}}/{{revision}}/deploy/operator.yaml :::
::::{caution}
{{productName}} deployment references its own image that it later runs alongside each ScyllaDB instance. Therefore, you have to also replace the image in the environment variable called SCYLLA_OPERATOR_IMAGE:
:::{code-block} yaml
:linenos:
:emphasize-lines: 16,19,20
apiVersion: apps/v1
kind: Deployment
metadata:
name: scylla-operator
namespace: scylla-operator
spec:
template:
# ...
spec:
# ...
containers:
- name: scylla-operator
# ...
image: docker.io/scylladb/scylla-operator:1.14.0@sha256:8c75c5780e2283f0a8f9734925352716f37e0e7f41007e50ce9b1d9924046fa1
env:
# ...
- name: SCYLLA_OPERATOR_IMAGE
value: docker.io/scylladb/scylla-operator:1.14.0@sha256:8c75c5780e2283f0a8f9734925352716f37e0e7f41007e50ce9b1d9924046fa1
:::
The {{productName}} image value and the SCYLLA_OPERATOR_IMAGE shall always match.
Be careful not to use a rolling tag for any of them to avoid an accidental skew!
::::
:::{code-block} shell
kubectl wait --for='condition=established' crd/scyllaclusters.scylla.scylladb.com crd/nodeconfigs.scylla.scylladb.com crd/scyllaoperatorconfigs.scylla.scylladb.com crd/scylladbmonitorings.scylla.scylladb.com
kubectl -n=scylla-operator rollout status --timeout=10m deployment.apps/{scylla-operator,webhook-server}
for i in {1..30}; do { kubectl -n=cert-manager get secret/cert-manager-webhook-ca && break; } || sleep 1 done :::
:::{caution} The following step heavily depends on the platform that you use, the machine type, or the options chosen when creating a node pool.
Please review the NodeConfig and adjust it for your platform! :::
:::::{tab-set}
::::{tab-item} GKE (NVMe) :::{code-block} shell :substitutions: kubectl -n=scylla-operator apply --server-side -f=https://raw.githubusercontent.com/{{repository}}/{{revision}}/examples/gke/nodeconfig-alpha.yaml ::: ::::
::::{tab-item} EKS (NVMe) :::{code-block} shell :substitutions: kubectl -n=scylla-operator apply --server-side -f=https://raw.githubusercontent.com/{{repository}}/{{revision}}/examples/eks/nodeconfig-alpha.yaml ::: ::::
::::{tab-item} Any platform (Loop devices) :::{caution} This NodeConfig sets up loop devices instead of NVMe disks and is only intended for development purposes when you don't have the NVMe disks available. Do not expect meaningful performance with this setup. ::: :::{code-block} shell :substitutions: kubectl -n=scylla-operator apply --server-side -f=https://raw.githubusercontent.com/{{repository}}/{{revision}}/examples/generic/nodeconfig-alpha.yaml ::: ::::
:::::
:::{note} Performance tuning is enabled for all nodes that are selected by NodeConfig by default, unless opted-out. :::
:::{code-block} shell
kubectl wait --for='condition=Reconciled' --timeout=10m nodeconfigs.scylla.scylladb.com/scylladb-nodepool-1 :::
:::{code-block} shell :substitutions:
kubectl -n=local-csi-driver apply --server-side -f=https://raw.githubusercontent.com/{{repository}}/{{revision}}/examples/common/local-volume-provisioner/local-csi-driver/{00_namespace,00_scylladb-local-xfs.storageclass,10_csidriver,10_driver.serviceaccount,10_provisioner_clusterrole,20_provisioner_clusterrolebinding,50_daemonset}.yaml :::
:::{code-block} shell
kubectl -n=local-csi-driver rollout status --timeout=10m daemonset.apps/local-csi-driver :::
:::{include} ../.internal/manager-license-note.md :::
:::::{tab-set}
::::{tab-item} Production (sized) :::{code-block} shell :substitutions: kubectl -n=scylla-manager apply --server-side -f=https://raw.githubusercontent.com/{{repository}}/{{revision}}/deploy/manager-prod.yaml ::: ::::
::::{tab-item} Development (sized) :::{code-block} shell :substitutions: kubectl -n=scylla-manager apply --server-side -f=https://raw.githubusercontent.com/{{repository}}/{{revision}}/deploy/manager-dev.yaml ::: ::::
:::::
:::{code-block} shell
kubectl -n=scylla-manager rollout status --timeout=10m deployment.apps/scylla-manager :::
Now that you've successfully installed {{productName}}, it's time to look at how to run ScyllaDB.