- Same AZ (snapshot can be used to copy EBS to another region/AZ)
- One instance at a time (can be detached / reattached to another)
- General Purpose SSD (gp2) - general purpose SSD, performance defined in IOPS
- Provisioned IOPS SSD (io1) - fastest
- Throughput Optimized HDD (st1) - low cost, HDD, performance defined by through-put, ideal for large, sequential work
- Cold HDD (sc1) - low cost, performance defined by through-put, for large, sequential, cold-data workloads
like raw, unformatted block devices
- select a customer master key (CMK) that you created separately using AWS KMS
- stored in S3 (can optionally specify a different AZ)
- supports multi-volume snapshots
- can be copied, encrypted and shared
- are incremental backups
- only changed block from last snapshot saved
- deletion process:
- only data unique to that snapshot deleted
- only need to save latest snapshot to restore (snapshot has all information)
- can be used to create new EBS volume
- automatically encrypted for encrypted volumes
- can periodically create snapshots - successive snapshots only charged by the difference
- Amazon Data Lifecycle Manager (Amazon Data Lifecycle Manager) to automate the creation, retention, and deletion of snapshots
- restore
- background lazy loading (can attach immediately after restore without waiting for complete data transfer)
- Fast Snapshot Restore - create fully restored volume
- encrypted snapshot owned by others - must first copy to own account (and own)
- Attach - can attach to running instance
- EBS volumes are exposed as NVMe block devices on Nitro-based instances. The device names are /dev/nvme0n1, /dev/nvme1n1, and so on.
- Format & mount - OS specific
- Status Monitoring (generate events)
- Every 5 minutes tests volume
- ok -> insufficient data (still testing) -> I/O disabled (something wrong) -> impaired (still problem, IO disabled)
- Auto-Enable IO : don't auto disable IO
- Performance status checking (only available for provisioned IO type)
- Impaired volume handling
- Opt1: disable applications (that uses volume); enable IO & check consistency with same instance (fsck); may contact support
- Opt2 (may lose writes still in memory): detach (force); enable IO; attach to another instance; check (fsck)
- Detach - unmount from OS first
Dynamic configuration change
- increase size (be aware of OS/file system constraints)
- IOPS capacity
- volume type
- Some EC2 instance types are EBS–optimized (or able to)
- Can be enabled at launch / running