Requirements
Required roles
For Cloudera Data Hub:
- DataHubAdmin, or
- Owner on the Data Hub, or
- Equivalent environment-level role
For Data Lake:
- Owner on the environment, or
- EnvironmentAdmin
Prerequisites
- The cluster must be in the Running state. The restart-runtime-services does not apply to stopped, failed, or upgrading clusters.
- Cloudera Manager (CM) must be reachable on the cluster.
- Any other flow on the same cluster (upgrade, scale, repair, another restart) must have finished. CM will refuse concurrent restart commands.
Flags
Both cdp datahub restart-runtime-services and cdp datalake restart-runtime-services accept the same two flags:
| Flag | Function |
|---|---|
| --rolling-restart | Instruct CM to restart role instances in batches rather than all at once. Minimizes downtime, but takes considerably longer. |
| --only-restart-stale-services | Instructs CM to only restart services currently marked as stale. Services already in FRESH state are skipped. |
You can combine these flags, use either one on its own, or ignore them.
Recommended combinations
| Situation | Recommended flags |
|---|---|
|
Stale configuration warning on a production Cloudera Data Hub or Data Lake |
--rolling-restart --only-restart-stale-services |
|
Full CM restart during a planned maintenance window; downtime acceptable |
No flags |
|
Restart everything but keep the cluster online as much as possible (for example, after a broad change) |
--rolling-restart |
|
Any use with --only-restart-stale-services without --rolling-restart |
This is not recommended because the stale filter is ignored in normal restart mode. |
