Upgrade a Silo Tenant
Version scope: commands and illustrative output target Server 20261109 (RELEASE.2026-11-09T00-00-00Z). See downloads and the release upgrade requirements.
The following procedures upgrade a single Silo Tenant using either Kustomize or Helm, and also cover standalone helm/silo chart 7.0.4 upgrades. Test the exact server image, Operator/chart version, and rollback procedure in a non-production cluster first.
Keep the server image on pgsty/silo and use only a tag or digest published on the Silo download page. The upstream Tenant defaults use a MinIO image. SILO disables in-place self-update; MINIO_UPDATE=off does not re-enable it. Upgrade by changing the Tenant or StatefulSet image version.
Review historical upstream migration requirements when upgrading very old deployments before switching to SILO.
Legacy AD/LDAP migrations
For Tenants running an upstream MinIO image older than RELEASE.2024-03-30T09-41-56Z with AD/LDAP enabled, read the upstream release notes for RELEASE.2024-04-18T19-09-19Z and complete those migration steps before upgrading to SILO.
Before You Upgrade to 20261109
Complete the following checks before upgrading any Kubernetes Tenant or standalone helm/silo release to SILO 20261109 (RELEASE.2026-11-09T00-00-00Z):
- Use a stop-all restart for 18+-pod single-pool layouts. A single pool whose erasure sets do not all share the same pods/hosts (18 or more pods with automatic layout) requires a drained stop-all restart—scale the StatefulSet to
0and confirm all old pods terminate before starting20261109pods. Do not run a rolling StatefulSet update on that topology, because old and new single-pool delete and copy lock domains (#298) do not exclude each other across sets. - Pause tier restores during mixed versions. Upgraded pods return
503 Service UnavailabletoPOST ?restorewhile any peer runs an older version or is unreachable. Pause new restore requests until all pods run20261109. Full multi-pool lifecycle and S3 bulk-delete Object Lock protections also require every serving pod to run20261109. - Drain batch jobs before changing versions. Stop batch submissions and drain or cancel all running, pending, and paused jobs before upgrading. Unfinished legacy jobs must be resubmitted after all pods upgrade, whereas completed jobs stay completed. Do not downgrade while unfinished
20261109jobs remain. Follow the batch upgrade and rollback checklist. - Preserve headless Service bootstrap and run a data-path check. Keep
publishNotReadyAddresses: trueon the peer discovery headless Service (*-hl) so pods discover peers during startup. After pods restart, run a bounded signed S3 write and readback check through each pod before resuming application traffic;readyandclusterprobes alone do not prove every remote drive is reconnected. See Startup Readiness Window. - Pin the image tag when reusing Helm values. Standalone
helm/silochart7.0.4(#296) deploys SILO andmcliRELEASE.2026-11-09T00-00-00Zby default, buthelm upgrade --reuse-valueskeeps the previously storedimage.tag. Pass--set image.tag=RELEASE.2026-11-09T00-00-00Z(or--set tenant.image.tag=RELEASE.2026-11-09T00-00-00Zfor the Operator Tenant chart) explicitly whenever you upgrade with--reuse-values.
Upgrade a Tenant using Kustomize
The following procedure upgrades a MinIO Tenant using Kustomize and the kubectl CLI. If you deployed the Tenant using Helm, use the Upgrade the Tenant using the MinIO Helm Chart procedure instead.
If the tenant was deployed with Operator Console, create a base configuration file before upgrading. If the tenant was deployed with Kustomize, use your existing kustomization files as the base configuration.
Choose a tab below depending on how the tenant was deployed:
-
Create the base configuration file:
-
In a working directory, save the current Tenant configuration to a file using
kubectl get:Replace
my-tenantandmy-tenant-nswith the name and namespace of the Tenant to upgrade.Edit the file to remove cluster-generated metadata fields:
creationTimestamp:resourceVersion:uid:selfLink:(if present)
For example, remove those fields from
metadata: -
In the same directory, create a
kustomization.yamlfile with contents resembling the following:If you used a different filename for the
kubectl getoutput in the previous step, replacemy-tenant-base.yamlwith the name of that file.
-
- Upgrade the tenant using the
kustomizationfiles from the original deployment as the base configuration. If you no longer have these files, follow the instructions in the Operator Console-Deployed Tenant tab.
- Create an
upgrade-minio-tenant.yamlfile with contents resembling the following:
This file instructs Kustomize to upgrade the tenant using the specified image. The name of this file, upgrade-minio-tenant.yaml, must match the patches.path filename specified in the kustomization.yaml file created in the previous step.
Replace my-tenant and my-tenant-ns with the name and namespace of the Tenant to upgrade. Replace the sample image tag only with a published Silo release that you have validated.
Alternatively, you can update the base configuration directly, according to your local procedures. Refer to the Kustomize Documentation for more information.
- From the same directory as the above files, apply the updated configuration to the Tenant with
kubectl apply:
The output resembles the following:
Upgrade the Tenant using the MinIO Helm Chart
This procedure upgrades an existing MinIO Tenant using Helm Charts.
If you deployed the Tenant using Kustomize, use the Upgrade a Tenant using Kustomize procedure instead.
-
Verify the existing Silo Tenant installation.
Use
kubectl get all -n TENANT_NAMESPACEto verify the health and status of all Tenant pods and services.Use the
helm listcommand to view the installed charts in the namespace:The result should resemble the following:
-
Update the Operator Repository
Use
helm repo update minio-operatorto update the MinIO Operator repo. If you set a different alias for the MinIO Operator repository, specify that to the command. You can usehelm repo listto review your installed repositories.Use
helm searchto check the latest available chart version after updating the Operator Repo:The response should resemble the following:
The
minio-operator/minio-operatoris a legacy chart and should not be installed under normal circumstances. -
Preserve and review the Tenant values
Export the release’s current user-supplied values, then verify that the file retains all topology, storage, TLS, credentials, and scheduling settings:
Set
tenant.image.repositorytodocker.io/pgsty/silo, pintenant.image.tagtoRELEASE.2026-11-09T00-00-00Z(or another tested published Silo release), and ensuretenant.envincludesMINIO_UPDATE=off. Never allow a chart upgrade to restore the upstream image default. -
Run the pinned
helm upgradePin the chart version separately from the Silo server image and pass the reviewed values file:
For standalone
helm/silochart7.0.4deployments (without MinIO Operator), chart7.0.4setsimage.tagandmcImage.tagtoRELEASE.2026-11-09T00-00-00Zby default (#296). When runninghelm upgrade --reuse-values, explicitly pass--set image.tag=RELEASE.2026-11-09T00-00-00Zso Helm does not retain the previously stored tag.The command returns success and increments
REVISION. -
Validate the Tenant Upgrade
Check that all services and pods are online, confirm the running image digest, and run a bounded signed S3 write and readback check through each pod before resuming application traffic; see Startup Readiness Window.