Question 1

What's the difference between gp2 and gp3 EBS volumes?

Accepted Answer

gp3 is the current generation general purpose SSD volume type and is roughly 20% cheaper per GB than gp2. gp3 ships with a flat 3,000 IOPS and 125 MB/s throughput at any size, while gp2 scales IOPS with size and only reaches 3,000 IOPS at 1,000 GB. For most workloads gp3 is a strict upgrade.

Question 2

How do you find gp2 EBS volumes across multiple AWS accounts?

Accepted Answer

Connect each account through a read-only cross-account IAM role and list volumes in every enabled region. Infralyst does this automatically across your connected accounts, returning every gp2 volume with its size, attachment state, and estimated monthly saving from a gp3 migration.

Question 3

Does AWS Trusted Advisor migrate gp2 volumes to gp3 automatically?

Accepted Answer

No. Trusted Advisor flags gp2 volumes as a cost optimization check but stops at the recommendation. Someone still has to open the Terraform file, change the type, and ship a PR. Infralyst closes that loop by generating the Terraform pull request for every gp2 volume it finds.

Question 4

Is there any workload where gp2 is still a better choice than gp3?

Accepted Answer

Not for general purpose workloads. gp3 matches gp2 on durability and block interface, costs 20% less per GB, and has a higher baseline IOPS floor. The only reason to keep gp2 is inertia. If you need more than 3,000 IOPS or 125 MB/s on gp3, provision them explicitly on the volume.

Question 5

Does migrating an EBS volume from gp2 to gp3 cause downtime?

Accepted Answer

No. AWS performs the type change as a live ModifyVolume operation. The volume stays attached, no detach, no remount, no snapshot, no reboot. Your application keeps running throughout. Infralyst only generates the Terraform PR, your team applies it on your normal deploy cadence.

Question 6

What happens to volumes that need more than 3,000 IOPS or 125 MB/s?

Accepted Answer

gp2 derives IOPS and throughput from volume size, so a large gp2 volume can have a baseline above gp3's defaults. Infralyst computes that baseline per volume from its size and region, and if it exceeds 3,000 IOPS or 125 MiB/s the generated PR sets explicit iops and throughput on the aws_ebs_volume resource to preserve it.

Question 7

How does Infralyst calculate the exact savings for a gp2 to gp3 migration?

Accepted Answer

Infralyst reads each volume's size and region, then prices gp2 and gp3 storage against the current AWS Price List for that region. The delta is the monthly saving. Every number you see in the dashboard, the Slack notification, and the PR body is computed per volume, not estimated from a flat percentage.

Question 8

What if EBS volumes are defined inside a shared Terraform module?

Accepted Answer

The Infralyst AI agent reads your Terraform repository, locates the aws_ebs_volume resource whether it lives at the root or inside a module, and changes the type at the right level. If the type comes from a shared variable that other volumes depend on, the agent overrides the type on the target resource instead of mutating the shared variable.

Question 9

What IAM permissions does Infralyst need to analyze EBS volumes?

Accepted Answer

A read-only cross-account IAM role. Infralyst reads EBS volume metadata (type, size, attachment state) through the EC2 API. It has no permissions to create, modify, delete, or detach volumes. Changes reach your infrastructure only through Terraform pull requests that your team reviews and merges.

Question 10

Is it safe to run a gp2 to gp3 migration on production databases?

Accepted Answer

Yes. gp3 has the same durability guarantees as gp2 and the ModifyVolume operation runs in the background while the volume stays mounted. For large volumes whose gp2 size-derived baseline exceeds gp3 defaults, the generated PR sets explicit iops and throughput so production databases don't lose IOPS or throughput in the migration. Infralyst generates one Terraform PR per volume, so you can stage the change on a non-critical volume first and merge the rest once you're comfortable.

Question 11

Could a gp2 to gp3 migration silently regress my volume's performance?

Accepted Answer

It can if you migrate a large gp2 volume to default gp3. gp2 scales IOPS at 3 per GiB and throughput up to 250 MiB/s with size, while gp3's defaults are flat at 3,000 IOPS and 125 MiB/s. Any gp2 volume above ~1 TiB or ~334 GiB has a baseline that exceeds those defaults. Infralyst's analysis computes the gp2 baseline per volume and, when it's higher than gp3 defaults, the generated PR sets explicit iops and throughput on the gp3 resource so the gp3 volume keeps the same baseline.

Question 12

Can I roll back a gp3 migration if something goes wrong?

Accepted Answer

Yes. Revert the Terraform change to flip the type back to gp2 and AWS runs another live ModifyVolume in reverse. No data loss, no downtime, no snapshot restore. In practice there's almost never a reason to roll back, gp3 is strictly cheaper and faster at baseline than gp2.

Resource	Size	Current	Recommended	Savings
postgres_dataus-east-1	500 GB	gp2	gp3	$10/mo	Generate PR
logs_archiveeu-west-1	2000 GB	gp2	gp3	$44/mo	Generate PR
app_stateus-east-1	100 GB	gp2	gp3	$2/mo	Generate PR

EBS Volume Optimization

You're paying 20% more for
EBS storage than you need to

How the analysis works

1. Collect volume inventory

2. Identify gp2 volumes

3. Generate the Terraform PR

Why gp3 beats gp2

20% cheaper per GB

Better baseline performance

AWS recommends it

Thresholds & methodology

IOPS baseline

Throughput baseline

What the PR looks like

Migrate EBS Volume logs_archive: gp2 → gp3 (-$44/mo)

Trigger PRs directly from Slack

Safety & guardrails

Zero downtime

Baseline preserved

Easy to revert

PR-only changes

EBS Optimization FAQ

Learn more about EBS optimization

How to Migrate EBS Volumes from gp2 to gp3

Start optimizing your EBS volumes