brandon/oam
1
0
Fork 0
mirror of https://gitea.com/mcereda/oam.git synced 2026-02-09 05:44:23 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
a1feef2e9150ff7b8320d00121147c8e152c186c
oam/knowledge base/cloud computing/aws/ebs.md
Michele Cereda 99709e9d0d chore(aws/ebs): review extending volumes
2025-08-21 21:24:50 +02:00

17 KiB
Raw Blame History

Elastic Block Store

Persistent block storage for EC2 Instances.

  1. TL;DR
  2. Volume types
  3. Snapshots
  4. Encryption
  5. Operations
    1. Increase disks' size
    2. Migrate gp2 volumes to gp3
  6. Further readings
    1. Sources

TL;DR

Real world use cases 
# Clean up unused volumes.
aws ec2 describe-volumes --output 'text' --filters 'Name=status,Values=available' \
  --query "Volumes[?CreateTime<'2018-03-31'].VolumeId" \
| xargs -pn '1' aws ec2 delete-volume --volume-id

# Check state of snapshots.
aws ec2 describe-snapshots --snapshot-ids 'snap-0123456789abcdef0' \
  --query 'Snapshots[].{"State": State,"Progress": Progress}' --output 'yaml'

# Wait for snapshots to finish.
aws ec2 wait snapshot-completed --snapshot-ids 'snap-0123456789abcdef0'

Every EBS volume is local to, and available in, a single Availability Zone.

Multiple EBS volumes can be attached to a single instance as long as both the volumes and the instance are in the same Availability Zone.

Depending on the volume and instance types, multiple instances can mount a single volume at the same time.

Volumes can have their size increased, but not reduced.
The volumes' filesystem is not automatically extended during the change in size, and must be extended manually to take advantage of the change in size.
Linux-based instances should™ be able to take care of that automatically after reboot.

Volume costs depend on its type, provisioned size, IOPS and throughput.
Volumes are billed per-second increments, with a 60-seconds minimum period.
Refer Amazon EBS pricing

EBS volumes attached to instances based on the Nitro system are exposed as NVMe devices.

Usage monitoring is available via instance-level metrics on CloudWatch per operations (EBSReadOps and EBSWriteOps) and bytes transferred (EBSReadBytes and EBSWriteBytes).

Instances built on the AWS Nitro system (EBS-Optimized) are capable of bursting performance.
Instances from medium to 4xlarge provide burst balance metrics to give information about the percentage of I/O and bytes credits remaining in the respective burst buckets:

  • EBSIOBalance% monitors the instance's I/O burst bucket.
  • EBSByteBalance% monitors the instance's byte burst bucket.

Large block workloads usually do not drive enough IOPS to deplete EBSIOBalance%, depleting EBSByteBalance% instead.
Small block workloads usually drive higher IOPS than bytes/second, dropping EBSIOBalance% faster than EBSByteBalance%.

Instances can drive EBS burst performance as long as EBSIOBalance% and EBSByteBalance% are above 0%.
When the I/O activity is below the baseline rate, the burst buckets refill.

The refill rate for burst buckets is the difference between the baseline rate and the I/O activity.
In addition, burst buckets are topped off every 24 hours, allowing instances to have burst performance available for at least 30 minutes per day.

Refer Amazon EBS-optimized instance types for details about instances performances.
Refer Improving application performance and reducing costs with Amazon EBS-Optimized Instance burst capability for details about EBS balances.

Volume types

Refer Amazon EBS volume types.

gp3 gp2 io2 io1 st1 sc1
Class SSD SSD SSD SSD HDD HDD
Annual failure rate 0.1% - 0.2% 0.1% - 0.2% 0.001% 0.1% - 0.2% 0.1% - 0.2% 0.1% - 0.2%
Size 1 GiB - 16 TiB 1 GiB - 16 TiB 4 GiB - 64 TiB 4 GiB - 16 TiB 125 GiB - 16 TiB 125 GiB - 16 TiB
Max IOPS 16,000 16,000 256,000 64,000 500 250
Max throughput 1,000 MiB/s 250 MiB/s 4,000 MiB/s 1,000 MiB/s 500 MiB/s 250 MiB/s
Multi-attach No No Yes Yes No No
NVMe reservations No No Yes No No No
Bootable Yes Yes Yes Yes No No
Pricing Per-GB + Per-IOPS over 3,000 + Per-MB/s over 125 Per-GB Per-GB + Per-IOPS Per-GB + Per-IOPS Per-GB Per-GB

Billing is per-second increments, with a 60-seconds minimum period.

Pricing examples:

gp3, 200 GB, 3000 IOPS, 512 MiB/s, for 69h 54m 34s in a 30d month in Ireland 
Regional price (storage): $0.088/GB/month
Regional price (IOPS): $0.0055/IOPS/month over 3000
Regional price (throughput): $0.044/MB/s/month over 125

Seconds in a 30d month: 60s * 60m * 24h * 30d = 2592000s
Seconds of actual usage: 34s + ( 60s * 54m ) + ( 60s * 60m * 69h ) = 34s + 3240s + 248400s = 251674s

Storage costs: 200GB * $0.088/GB * ( 251674s / 2592000s ) = 200 * $0.088 * 0.09709645062 = $1.71
IOPS costs: ( 3000 - 3000 )IOPS * $0.0055/IOPS * ( 251674s / 2592000s ) = 0 * $0.0055 * 0.09709645062 = $0.00
Throughput costs: ( 512 - 125 )MB/s * $0.044/MB/s * ( 251674s / 2592000s ) = 387 * $0.044 * 0.09709645062 = $1.66

Total: $1.71 + $0.00 + $1.66 = $3.37

Snapshots

A volume's first snapshot is a complete snapshot of it, with all the volume's blocks being copied over.
All successive snapshots of the same volume are incremental, with only the changes being copied over.
Incremental snapshots are stored in EBS' standard tier.

Snapshots can be unbearably slow depending on the amount of data needing to be copied.
For comparison, the first snapshot of a standard 200 GiB gp3 volume took about 2h to complete.

Snapshots can be archived to save money should they not need frequent nor fast retrieval.
When archived, incremental snapshots are converted to full snapshots and moved to EBS' archive tier.

The minimum archival period is 90 days.
Archived snapshots deleted or permanently restored before the end of the minimum archival period are billed for the whole period.

When access to archived snapshots is needed, they need to be restored to the standard tier before use. Restoring can take up to 72h.

Lifecycle policies' targetTags attribute targets resources of the specified type in an OR fashion, not AND, meaning they will target all resources with at least one of the defined target tags.

Encryption

Refer How Amazon EBS encryption works.

One can encrypt both boot and data volumes.

At the time of writing, only symmetric keys are supported.

Volumes attached to supported instance types encrypt the following types of data:

  • Data at rest inside the volume.
  • Data moving between the volume and the attached instance.
  • Snapshots created from the volume.
  • Volumes created from said snapshots.

Volumes are encrypted with a AES-256 data key.
The key is:

  1. Generated by KMS.
  2. Encrypted by KMS with another KMS-managed key.
  3. Stored with the volume's information.

EBS automatically creates a unique AWS-managed key in each Region where one creates EBS resources, using the aws/ebs alias. EBS then uses this KMS key for encryption by default.
Alternatively, one can use a symmetric customer managed encryption key of one's own creation.

EC2 integrates with KMS to encrypt and decrypt EBS volumes in ways that differ depending on whether the original snapshot for encrypted volumes is itself encrypted or unencrypted.

The original snapshot is encrypted
  1. EC2 sends a GenerateDataKeyWithoutPlaintext request to KMS specifying the KMS key for volume encryption.
  2. If the volume is encrypted using the same key as the snapshot, KMS encrypts that key using that same data key as the snapshot.
    If the volume is encrypted using a different KMS key, KMS generates a new data key and encrypts it using the specified key. The encrypted data key is then sent to EBS to be stored with the volume metadata.
  3. When attaching the encrypted volume to an instance, EC2 sends a CreateGrant request to KMS to be allowed to decrypt the data key.
  4. KMS decrypts the encrypted data key and sends the decrypted data key to EC2.
  5. EC2 uses the plaintext data key in the Nitro hardware to encrypt disk I/O to the volume.
    The plaintext data key persists in memory as long as the volume is attached to the instance.
The original snapshot is not encrypted

  1. EC2 sends a CreateGrant request to KMS to be allowed to encrypt the volume that is being created from the snapshot.

  2. EC2 sends a GenerateDataKeyWithoutPlaintext request to KMS specifying the key chosen for volume encryption.

  3. KMS generates a new data key, encrypts it using the specified key, and sends the encrypted data key to EBS to be stored with the volume metadata.

  4. EC2 sends a Decrypt request to KMS to decrypt the encrypted data key, which it then uses to encrypt the volume's data.

  5. When attaching the encrypted volume to an instance, EC2 sends:

    1. A CreateGrant request to KMS to be allowed to decrypt the data key.
    2. A Decrypt request to KMS specifying the encrypted data key.

  6. KMS decrypts the encrypted data key and sends the decrypted data key back to EC2.

  7. EC2 uses the plaintext data key in the Nitro hardware to encrypt disk I/O to the volume.
    The plaintext data key persists in memory as long as the volume is attached to the instance.

When KMS keys become unusable, the effect is almost immediately subject to eventual consistency.
The key state of the impacted KMS keys change to reflect their new condition, and all requests to use those keys in cryptographic operations fail.

EC2 encrypts all I/O to and from attached volumes using the data key, not the KMS key itself.
There is no immediate effect on the EC2 instance or its attached EBS volumes when performing actions that make KMS keys unusable.

EBS removes data keys from the hardware when encrypted EBS volumes are detached from instances.
Attaching EBS volumes which data keys are encrypted with unusable KMS keys to EC2 instances will fail, because EBS will not be able to use the KMS keys to decrypt the data key used for the volume.
Make the KMS key usable again to be able to attach such EBS volumes.

Operations

Increase disks' size

Refer Modify an Amazon EBS volume using Elastic Volumes operations and How do I increase or decrease the size of my EBS volume?.

If an EC2 instance uses EBS volumes, a volume's size can be increased without needing to detach the volume first or restart the EC2 instance. This allows to continue using the EC2 instance while the changes take effect.

  1. Increase the volume's size:

    aws ec2 modify-volume --volume-id 'vol-0123456789abcdef0' --size '512'

# Check on the operation
aws ec2 describe-volumes-modifications --volume-ids 'vol-0123456789abcdef0' --query 'VolumesModifications[]'

  2. Extend the volume's partitions from inside the instance using it:

    lsblk
sudo growpart '/dev/nvme0n1' '1'  # nitro
sudo growpart '/dev/xvda' '1'     # xen

  3. Extend the volume's file system from inside the instance using it:

    sudo xfs_growfs -d '/'           # xfs
sudo resize2fs '/dev/nvme0n1p1'  # ext4 on nitro
sudo resize2fs '/dev/xvda1'      # ext4 on xen

After modifying a volume, one must wait at least six hours and ensure that the volume is in the in-use or available states before one can modify the same volume.

Modifying the volume can take from a few minutes to a few hours, depending on the changes being applied, and it does not always scale linearly.
A volume of 1 TiB in size can typically take up to six hours to be modified, but it could take 24 hours or longer in certain situations.

Migrate gp2 volumes to gp3

See also Hands-on Guide: How to migrate from gp2 to gp3 volumes and lower AWS cost.

It is strongly advised to take a snapshot of volumes before changing their type.

aws ec2 modify-volume --volume-id 'vol-0123456789abcdef0' --volume-type 'gp3'

# Do it for all volumes that are currently of type 'gp2'
aws ec2 describe-volumes --filters "Name=volume-type,Values=gp2" --query 'Volumes[].VolumeId' --output 'text' \
| xargs -pn '1' aws ec2 modify-volume --volume-type 'gp3' --volume-id

If changing the volume type from gp2 to gp3 without specifying IOPS or throughput performance, EBS automatically provisions either equivalent performance to that of the source gp2 volume, or the baseline gp3 performance, whichever is higher.

Further readings

Sources