Is the HA option for Cloud SQL also a matter of data redundancy?

I'm looking at the GCE "Cloud SQL" product, and specifically the MySQL flavour. I'm a bit confused about how this product operations in non-HA mode.

I understand that in HA mode, you have a standby replica ready to kick in if your primary instance becomes unavailable.

However, what happens if a non-HA instance becomes unavailable? Will GCE always, eventually at some point, bring it back up again? When it is brought up again, would worst case scenario then be that your data is 24 hours old (last daily backup)? Or is manual work actual required in this case; e.g you need to provision a new instance based on a backup yourself?

It appears to me that the non-HA mode has a guarantee of 99.95% availability. Is the HA-mode then what covers your back during the 0.05% remaining time?

Google says :

Cloud SQL automates all your backups, replication, encryption patches, and capacity increases—while ensuring greater than 99.95% availability, anywhere in the world.


Solution 1:

However, what happens if a non-HA instance becomes unavailable?

This means that your data will be unavailable until the instance become online in most of the cases this is caused by an user misconfiguration but in some cases you will covered after a review with support, with HA when the master become unavailable the standby enter online until the master returns (brief downtime maybe less than ~2min).

When it is brought up again, would worst case scenario then be that your data is 24 hours old (last daily backup)?

You always can use the backups to start a new instance, but if the instance become unhealthy and returns,this will not take an backup this will return with the latest data available before the downtime. Usually if the instance goes offline is not possible write over this.

Is the HA-mode then what covers your back during the 0.05% remaining time?

The SLA for both modes are the same, The sla covers full downtime for example on non HA instances your are covered since your instance become unavailable, but on HA instnace, you only are covered when Standby and Master are unavailable causing a full downtime.

For more information please check the CLOUD SQL SLA