Skip to main content
12.x upgrades in place from 11.x. Upgrade straight to the hub chart 12.0.2 (see the note below on 12.0.0); the exporter chart is 12.0.0. Both releases run on the same new-generation telemetry tables, so no ClickHouse reset, data migration, or datastore maintenance step is required. New PostgreSQL tables (the webhook secret store, CRD sync state, Guardian GitHub drafts) are created automatically when the new Apiserver starts.
Hub commands assume the Helm release is named metoro and runs in the metoro-hub namespace. Monitored-cluster exporter commands assume the Helm release is named metoro-exporter and runs in the metoro namespace. Adjust release names and namespaces if your installation uses different values.

Values Changes

12.0.0 removes the 11.x telemetry-table switches, because the hub now always writes and reads only the new-generation tables. Delete these keys from your hub values file if they are present:
  • newTelemetryTablesAcknowledged
  • ingester.tablesV3
  • ingester.tablesV3Only
  • ingester.attributesJsonTables
The chart ignores them if they are left in place, so forgetting one does not break the upgrade. New optional values, all off by default:
Do not upgrade to hub chart 12.0.0 or 12.0.1. 12.0.0 left the bundled PostgreSQL at its default max_connections of 100, which is too low for Temporal’s connection pools, so some Temporal pods crashloop with no usable database connection found. 12.0.1 raised max_connections to 400 (new value postgres.bundled.cluster.parameters) and sizes the Temporal pools to fit. If you use external PostgreSQL, make sure it allows at least 400 connections for Metoro. Separately, Guardian chat fails on 12.0.0 and 12.0.1 hubs served over plain HTTP; 12.0.2 fixes that. See Upgrade 12.0.0 or 12.0.1 to 12.0.2.
When ai.enabled=true, the Guardian sandbox pods now run with an Unconfined seccomp profile (and an Unconfined AppArmor profile unless agentSandbox.appArmorProfile is set), so the Guardian runner can create the user namespaces it uses for its own isolation. The pods still run as non-root with all capabilities dropped. If your cluster enforces Pod Security Admission restricted or a policy engine rule on the Agent Sandbox namespace, allow these profiles for the Guardian sandbox pods before upgrading.

Upgrade The Hub

Start from the values file used for the existing hub release. If you do not have it locally, export the user-supplied Helm values:
Remove the keys listed in Values Changes, then upgrade the hub chart to 12.0.2:

Check The Hub Rollout

If hub self-monitoring is enabled, check the bundled exporter rollout too:
If Metoro AI is enabled, check the Agent Sandbox components:

Upgrade Monitored-Cluster Exporters

Upgrading the hub repins future onboarding commands, but it does not upgrade exporter releases already installed on monitored clusters. The 12.0.0 exporter chart installs the new MetoroClusterIngestionRule CRD, adds spec.tags to the MetoroAlert and MetoroDashboard CRDs and the Secret reference fields to MetoroWebhook, so upgrade every monitored cluster before using those features there. Upgrade the hub first, then repeat this section for every monitored cluster that has metoro-exporter-onprem installed. Start from the values file used for the existing exporter release. If you do not have it locally, export the user-supplied Helm values:
Upgrade the exporter chart to 12.0.0:

Check Monitored-Cluster Rollouts

If ServiceMonitor scraping is enabled, check the collector and target allocator rollouts too:
Expected outcome:
  • Apiserver, Ingester, and (when enabled) the Guardian runner roll to the onprem-12.0.2 images, and the bundled PostgreSQL restarts once (rolling, with a switchover) to apply max_connections.
  • Hub self-monitoring rolls to the 12.0.0 exporter chart and images when enabled.
  • Existing monitored-cluster exporter releases roll to the 12.0.0 exporter chart and images after their Helm upgrades, and the MetoroClusterIngestionRule CRD exists on each cluster.
  • Existing PostgreSQL, ClickHouse, Temporal, and telemetry data remain in place.
  • Future monitored-cluster onboarding commands reference metoro-exporter-onprem 12.0.0.

Rolling Back

11.0.0 and 12.0.0 use the same telemetry tables, but 12.0.0 also updates ClickHouse view definitions and PostgreSQL schema on startup. Contact Metoro support before rolling a hub back to 11.0.0.