- Newest
- Most votes
- Most comments
Thanks Sean. In terms of the 2nd option, it looks like passwords can't be replicated. so how does that factor in? Will the user have to reset the password during a Disaster and reset it again when the primary comes back online?
While I cannot answer the question on when multi-region support is arriving I can give some suggestions that I have seen people use. I don't know the use case you have (identity pools, user pools, or both) so I will talk about user pools since that is the common use case.
The underlying database used by Cognito is highly resilient and I would expect it to survive a regional failure. However, it would be unavailable for the duration of a regional failure and that may well be unacceptable. If this is the case, and until multi-region support arrives, I would suggest the following:
First. Don't store user credentials in the Cognito database. Instead federate via SAML as this removes the biggest issue with single region use, the inability to back up user credentials. I prefer this anyway as I really want a single database of record for all my users and that often already exists.
Second. Design for failure. Assume that the region will fail and either build out parallel Cognito infrastructure in another region, or have the Cloudformation to do so ready to go. This is a cost vs. RTO trade off as to whether it is better to have warm or cold standby in your use case.
Finally. If at all possible store only transient session data in the database. Ideally if the region fails then you should be able to cold start in another region with minimal impact. If the data needs to persist then run regular backup jobs to walk the database and copy the data out according to your RPO.
We experience the same issue and got stuck solving it. But for us is also important synchronization of user pool app clients across the regions. Seems like we have to completely rid of Cognito from the solution and secure our API with API keys
Relevant content
- asked 4 months ago
- AWS OFFICIALUpdated 2 years ago
- AWS OFFICIALUpdated a year ago
- AWS OFFICIALUpdated 2 years ago
- AWS OFFICIALUpdated a year ago
Thanks Sean. In terms of the 2nd option, it looks like passwords can't be replicated. so how does that factor in? Will the user reset the password during a Disaster and reset it again when the primary comes back online?