Durability measures how likely stored data is to remain intact and recoverable over time despite failures. It differs from availability: a temporarily inaccessible EBS volume can return online with its data intact. section 13.2.3
A single EBS volume is confined to one Availability Zone. An outage can prevent a replacement instance in another zone from accessing it, even when the data has not been lost. section 13.2.3
Common ways to improve recovery and reduce data-loss risk include:
- Move state to managed, multi-zone services such as RDS, DynamoDB, EFS, or S3. section 13.2.3
- Create regular EBS snapshots; snapshots are stored on S3 and can be accessed across Availability Zones. section 13.2.3
- Use AMIs when the EBS volume is an ECS instance's root volume. section 13.2.3
- Use distributed storage such as GlusterFS, DRBD, or MongoDB when application requirements demand it. section 13.2.3
Recovering Jenkins state
Jenkins stores data directly on disk, so an object or database service is unsuitable for its file-level state. EFS provides the shared file storage needed across instances and Availability Zones. section 13.2.3
