You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In a single-master Kubernetes cluster: After the apiserver master node exits, Lorry's lease judgment becomes entirely unavailable.
In a three-master Kubernetes cluster: After the apiserver master node exits, testing reveals that Lorry's lease judgment might result in errors, causing the high-availability PostgreSQL cluster to re-elect a new primary.
The PG switchover issue arises from Patroni when the API server is down, and Lorry do not demote the primary in case of a failed API server access. To address this problem, there are two possible approaches:
Implement high availability for the API server and perform upgrades one by one.
If the PG cluster has only one replica, disable the Patroni high availability functionality.
Consider the following scenarios:
In a single-master Kubernetes cluster: After the apiserver master node exits, Lorry's lease judgment becomes entirely unavailable.
In a three-master Kubernetes cluster: After the apiserver master node exits, testing reveals that Lorry's lease judgment might result in errors, causing the high-availability PostgreSQL cluster to re-elect a new primary.
code ref:
kubeblocks/pkg/lorry/dcs/k8s.go
Line 299 in 2d82e5a
The text was updated successfully, but these errors were encountered: