As indicated in the CRD reference the replication.mode async option is the default and enables the asynchronous Postgres replication mode. This is the most common way of creating a replica in a Postgres cluster.
Setting up the StackGres Cluster with asynchronous replica members is quite straightforward. In the Cluster Creation section, the example used the default way, async.
Nevertheless, the next box highlight the SGCluster CRD again:
cat << EOF | kubectl apply -f -
apiVersion: stackgres.io/v1
kind: SGCluster
metadata:
namespace: failover
name: cluster
spec:
postgres:
version: '16.1'
instances: 3
sgInstanceProfile: 'size-s'
pods:
persistentVolume:
size: '10Gi'
configurations:
sgPostgresConfig: 'pgconfig1'
sgPoolingConfig: 'poolconfig1'
prometheusAutobind: true
nonProductionOptions:
disableClusterPodAntiAffinity: true
replication:
mode: async
role: ha-read
syncInstances: 1
EOF
The result will be the next:
$ kubectl -n failover exec -it cluster-0 -c patroni -- bash - patronictl list
+ Cluster: cluster (7369933339677233777) +-----------+----+-----------+
| Member | Host | Role | State | TL | Lag in MB |
+-----------+------------------+---------+-----------+----+-----------+
| cluster-0 | 10.244.0.8:7433 | Leader | running | 1 | |
| cluster-1 | 10.244.0.10:7433 | Replica | streaming | 1 | 0 |
| cluster-2 | 10.244.0.12:7433 | Replica | streaming | 1 | 0 |
+-----------+------------------+---------+-----------+----+-----------+
Maybe the variable syncInstances caught your attention. As shown above the cluster is composed by 1 Leader and 2 replicas using asynchronous replication. Therefore, we could think that the variable should be set to 0 instances, but the documentation confirms that the variable will take effect only if we enable synchronous replication. Please, be aware of this if there are plans to update the replication mode from async to sync.
Nevertheless, an example is included to demonstrate that updating the variable is harmless if sync mode is not enabled:
cat << EOF | kubectl apply -f -
apiVersion: stackgres.io/v1
kind: SGCluster
metadata:
namespace: failover
name: async-cluster
spec:
postgres:
version: '16.1'
instances: 3
sgInstanceProfile: 'size-s'
pods:
persistentVolume:
size: '10Gi'
configurations:
sgPostgresConfig: 'pgconfig1'
sgPoolingConfig: 'poolconfig1'
prometheusAutobind: true
nonProductionOptions:
disableClusterPodAntiAffinity: true
replication:
mode: async
role: ha-read
syncInstances: 3
EOF
sgcluster.stackgres.io/async-cluster created
kubectl -n failover exec -it async-cluster-0 -c patroni -- bash - patronictl list
+ Cluster: async-cluster (7369943621678699243) +-----------+----+-----------+
| Member | Host | Role | State | TL | Lag in MB |
+-----------------+------------------+---------+-----------+----+-----------+
| async-cluster-0 | 10.244.0.14:7433 | Leader | running | 1 | |
| async-cluster-1 | 10.244.0.16:7433 | Replica | streaming | 1 | 0 |
| async-cluster-2 | 10.244.0.18:7433 | Replica | streaming | 1 | 0 |
+-----------------+------------------+---------+-----------+----+-----------+