1ctl valkey create
Create a private Valkey service with automatic placement or select a specific eligible machine.
1ctl valkey create <name>Arguments
Section titled “Arguments”| Argument | Required | Variadic |
|---|---|---|
name |
Yes | No |
| Flag | Type | Required | Default | Description |
|---|---|---|---|---|
--machine-id |
string | No | — | Specific eligible machine; automatic placement when omitted |
--topology |
string | No | "standalone" |
Topology: standalone or replicated |
--instances |
integer | No | 0 |
Total instances (default: 1 standalone, 3 replicated) |
--persistence |
boolean | No | true |
Persist data on retained volumes |
--storage-size |
string | No | "8Gi" |
Data volume size |
--storage-class |
string | No | — | Kubernetes storage class (default: auto-detect) |
--cpu-request |
string | No | "250m" |
CPU request per instance |
--cpu |
string | No | "500m" |
CPU limit per instance |
--memory-request |
string | No | "512Mi" |
Memory request per instance |
--memory |
string | No | "1Gi" |
Memory limit per instance |
--append-only |
boolean | No | true |
Enable AOF durability |
--append-fsync |
string | No | "everysec" |
AOF fsync policy: always, everysec, or no |
--maxmemory-policy |
string | No | "allkeys-lru" |
Valkey eviction policy |
--maxmemory-percent |
integer | No | 75 |
Memory limit percentage available to Valkey data |
--metrics |
boolean | No | true |
Enable Prometheus metrics exporter |
--help, -h |
boolean | No | — | Show help |
Examples
Section titled “Examples”Create a service using the generated defaults:
1ctl valkey create sessionsSelect a specific machine:
1ctl valkey create sessions-local \ --machine-id 89905f5f769867452a7bd6c7505ab34dPlacement behavior
Section titled “Placement behavior”Automatic placement checks both durable machine records and live Kubernetes Nodes. A machine must be Ready, schedulable, monetized, verified, reconciled as rentable capacity, and open for marketplace admission. The selected ID is persisted with the storage configuration.
Reconciliation requires the same machine and does not silently relocate data.
Persistent standalone services use Recreate rollout strategy to detach their
ReadWriteOnce volume before starting a replacement pod.
Replicated topology currently places every instance on that same selected
machine. It provides a read-only replica endpoint, not machine-level high
availability or automatic promotion.