Yva.ai backup policy

Yva.ai automatically performs data backup of the main databases and internal system settings is performed using the internal functionality. By default, the backup frequency is 1 time per day, the storage period of each backup is 4 days (after 4 days the backup is destroyed), and the period can be increased at the client's request. The backup mechanism is automated for Self-hosted solutions, Azure Self-hosted solutions, and Public Cloud installations with 1500 or more connected employees.

For Azure self-hosted solutions, we recommend an additional configuration of the snapshots of virtual machines creation using MS Azure. The recommended frequency is 1 time per day, the storage period is 7 days.

For Self-hosted solutions, we also recommend an additional configuration of backups in accordance with the company's policies.

In case of an incident:

  • if company loses the main databases (except for the report database), corresponding employee will need to restore them manually, this process will not be noticed by users, however collection of new passive data and/or sending of new surveys can be disabled for the period of recovery.
  • if company loses the report database, the platform will be completely inaccessible to users. The standard recovery time is up to 24 hours.
  • if company loses non-core databases, the system will automatically restore them in the background. Non-core databases include databases that contain data that can be recovered from a collaboration sources.

Yva.ai tests the data recovery from backups periodically during major releases. The data is deployed on additional servers, unnoticed by users.

Additionally, for configurations with 1500 or more connected employees, the main databases are mirrored using the Mongo Replica Set in order to provide system Uptime (sending surveys, displaying dashboards, etc.). A configuration consisting of Primary and Secondary servers and Arbiter is used. If the Primary or Secondary Server fails completely, the remaining active server will take over the entire current load for the recovery period. Users will not notice this situation.