Reputation: 13
Jan 2018 Spanner was down for 41 minutes in a region. What’s the alternative if Spanner goes down for 6 hours (assuming the worst case scenario) in a region.
Upvotes: 0
Views: 275
Reputation: 2324
There are effectively 2 alternatives:
Wait for the outage to recover...
Or
Import your last export (which you stored in a multi-region cloud storage bucket) into a new instance in another region.and point your application at that new instance.
This may take more time that the duration of the outage itself, and will require some complex re-syncing of data when the outage recovers.
There is also the preventative option of running a multi-region instance, but that can be very expensive, and has to be valued against the effective cost of an outage of your application.
Upvotes: 0
Reputation: 776
One of the specifications that you can see in the Cloud Spanner Overview is that is guaranteed max of 5 minutes of downtime (including planned downtime) on paper and in practice.
According to the Cloud Spanner SLA, there is specified a Monthly Uptime Percentage:
Multi-Regional Instance -> >=99.999%
Regional Instance -> 99.99%
Furthermore, in case Spanner was down for 6 hours in your region:
720 hours = total hours in a month
6 hours = Downtime suffered
Total Monthly Uptime Percentage = (720 - 6)/720 = 99.166%
As the Monthly Uptime Percentage is between 99.0% and 99.99%, and as specified in the documentation, you would get 10% of monthly bill for Cloud Spanner Regional that will be credited to future monthly bills.
Upvotes: 2