949237 - FXA DB needs to support an availability target of no more than 5 min outage or recovery (multi region / master)

Status: RESOLVED WONTFIX

(Cloud Services :: Server: Firefox Accounts, defect)

Description

[Edwin Wong [:edwong]]

FXA as a product needs to define the SLA around availability and recovery.  We need to have a DB topology that supports that roadmap.

Question include: 
1. does DB read/write downtime affect B2G users from using their phone?  
2. Does it affect where is my fox locating a device or locking a device remotely.  
3. How does downtime affect sync or possible storage / ReadItLater solutions.

Comment 1

[Chris Karlof [:ckarlof]]

For initial release, we are largely building a single AWS region deployment with somewhat manual recovery in the case of a region outage. My opinion is that this is acceptable for initial release. People have launched with less. However, I'd like explicit documentation of:

1) our availability goals with this setup
2) MTTR goals in the case of region outage, and response plan
3) a measurement plan to track whether we are meeting this goals
4) sign off from :mmayo and :lloyd on above items for initial release

Comment 2

[Edwin Wong [:edwong]]

removing lloyd from needinfo

Comment 3

[Ryan Kelly [:rfkelly]]

What we have is meeting our needs; Travis's team can revisit this in fresh bugs as opportunity arises