0

I use Azure Cosmos DB with multi-region write in two regions and I need to know what will happen if one of this regions fails (outages). Will it be data loss and if, yes, how much time will it take to recovery?

Maybe someone in his project met faced the situation when in Cosmos DB with multi-region write one of the region outages. Can someone share such experience? What scale of data loss or corruption do I risk?

Documentation says that that there is a possibility of data loss, but there is no clear information, how much data can be lost (such as a calculation formula or something similar), also there is no information about the time. "Given the internal Azure Cosmos DB architecture, using multiple write regions doesn't guarantee write availability during a region outage. The best configuration to achieve high availability during a region outage is single write region with service-managed failover." https://docs.microsoft.com/en-us/azure/cosmos-db/high-availability#what-to-expect-during-a-cosmos-db-region-outage

1 Answers1

0

This statement in the doc gives you a fairly succint answer:

Recently updated data in the failed region may be unavailable in the remaining active regions. Eventual, consistent prefix, and session consistency levels guarantee a staleness of <15mins. Bounded staleness guarantees less than K updates or T seconds, depending on the configuration. If the affected region suffers permanent data loss, unreplicated data may be lost.

If you are using eventual, consitent prefix or session consitency you could loose up to 15 mins worth of data, if you use boudless staleness then it will depend on your configuration.

Sam Cogan
  • 38,158
  • 6
  • 77
  • 113