Cloud SQL Zero downtime Maintenance

[chen levi]

Hi

looks like there is nothing we can do in order to enforce Zero downtime for our GCP Cloud SQL during maintenance window.

from what i've read seems like even adding Failover replicas will not help here as the replicas will have the same maintenance window, meaning they will both get down during maintenance.

also i assume read replicas also wont help here as read replicas are not ment for failover, although they have a different maintenance window (which can be anytime).

am i missing something?

thanks

Chen

[George (Cloud Platform Support)]

Short maintenance-related service interruptions do occur. You should design your applications to deal with situations when your instance is not accessible for short periods of time, such as in a maintenance shutdown. To minimize impact to your service, you can set a maintenance window to control when downtime occurs.

You can test the behavior of your application to a maintenance shutdown by restarting your instance, which has the same effect. In general, we recommend that you use only short-lived connections as well as use exponential back-off for retrying rejected connections. For more guidance see How should I manage connections?.

[Tenzing Namgyal Wangdi]

Hi, We have used an Active-Passive configuration of Cloud SQL postgres. In a sense it means that I have 2 Cloud SQL master in the same region. One is primary which is responsible for the entire application. It has some read replicas. But during maintenance of master 1 and read replicas, I have a HAProxy to switch to a backup server. Our data is high read/less write scenario for which the data syncing to the master 2 can be done with export from master 1 and import into backup master 2. Once the maintenance is over, HAProxy switches it back to master 1. So practically I have zero downtime. There are other partner solutions Google can offer so you can get in touch with your Technical Account Manager.