blog

AWS Extended Support: Finding the Charges and Deciding What to Do About Them

EKS, RDS, Aurora, ElastiCache and OpenSearch all auto-enroll old versions into paid extended support. How to find those charges on your bill, what's coming up next, and when paying the fee is the better call.

When a Kubernetes or database version on AWS reaches the end of standard support, AWS mostly doesn’t force an upgrade anymore. The resource keeps running as before, and AWS starts charging extra for extended support. It’s on by default, so the first sign is usually a new line on the bill.

We find these charges on most AWS bills we go through, often on resources nobody has looked at in a while. Here’s what each service charges, how to find it in your account, and how to choose between paying and upgrading.

EKS#

An EKS version gets 14 months of standard support, then 12 months of extended support. In extended support the control plane price goes from $0.10 to $0.60 per cluster per hour. That’s about $73 a month becoming $438, per cluster, before any nodes.

What to know:

  • It’s the default. Every cluster has an upgrade policy, and unless you changed it, it’s EXTENDED.
  • You can’t opt out once you’re in. The policy can only be changed while the cluster’s version is still in standard support. After that, the only way out is to upgrade.
  • STANDARD means auto-upgrade. With that policy, AWS upgrades the control plane to the next version when standard support ends. Your nodes stay on the old version until you upgrade them.
  • Extended support ends too. After the 12 months, AWS upgrades the control plane anyway. The end date is published well ahead, but the upgrade itself can happen any time after it, with no separate notice.

Dates from the EKS release calendar:

  • 1.31: extended support ends 26 November 2026.
  • 1.32 and 1.33: in extended support now, until 23 March 2027 and 29 July 2027.
  • 1.34: standard support ends 2 December 2026.

RDS and Aurora#

RDS Extended Support covers MySQL and PostgreSQL, on both RDS and Aurora. It’s enabled by default when you create or restore a database, and charges start the day after the major version’s end of standard support. It lasts up to three years. After that, AWS upgrades the database for you.

The price is per vCPU-hour, and it goes up in year 3:

  • AWS’s examples for US East use $0.100 per vCPU-hour in years 1 and 2, and $0.200 in year 3. Prices differ by region, so check the RDS pricing page for yours.
  • Multi-AZ standbys are charged too, and so are read replicas on the same version. A Multi-AZ instance with one standby pays twice.
  • Aurora Serverless is charged per ACU-hour instead of per vCPU.

An example, using the US East year 1 price: a db.r6g.xlarge has 4 vCPUs, so it pays $0.40 an hour, about $290 a month. With Multi-AZ it’s about $580. In year 3 both numbers double.

Dates:

  • RDS for MySQL 8.0: standard support ended on 31 July 2026. Instances still on 8.0 have been paying since 1 August.
  • RDS for MySQL 5.7: on year 3 pricing since 1 March 2026.
  • Aurora MySQL version 2 (5.7 compatible): in extended support since late 2024, and moves to year 3 pricing on 1 December 2026.
  • Aurora MySQL version 3 (8.0 compatible): standard support runs until 30 April 2028, so MySQL 8.0 on Aurora doesn’t cost extra yet.
  • RDS for PostgreSQL 13: paying since 1 March 2026.
  • RDS for PostgreSQL 14: standard support ends on 28 February 2027. PostgreSQL 12 moves to year 3 pricing the day after.

ElastiCache and OpenSearch#

ElastiCache for Redis OSS versions 4 and 5 have been in extended support since 1 February 2026. The charge is added on top of the on-demand node price: 80% in years 1 and 2, 160% in year 3. Reserved nodes don’t reduce it. Redis OSS 6 reaches end of standard support on 31 January 2027. Upgrading to Valkey or Redis OSS 7 stops the charge.

OpenSearch Service used to charge a flat rate for old versions, $0.0065 per Normalized Instance Hour in US East. Since 7 November 2026, the charge for the oldest versions equals the instance price, so instances on those versions cost double. That applies to Elasticsearch 1.5, 2.3, 5.1 to 5.5, 6.0 to 6.7 and 7.1 to 7.8, and to OpenSearch 1.0 to 1.2 and 2.3 to 2.9. Elasticsearch 5.6 stays on the flat rate. Storage pricing doesn’t change.

Finding it in your account#

Cost Explorer. Group by Usage Type and search the usage types for extended. Then group by service and region to see where the charges come from.

The commands below only see one region, the one your CLI is set to. Cost Explorer covers all of them, so use it to find the regions with charges, then run the commands in each one with --region.

EKS clusters, with version and upgrade policy:

sh
for c in $(aws eks list-clusters --query 'clusters[]' --output text); do
  aws eks describe-cluster --name "$c" \
    --query 'cluster.[name,version,upgradePolicy.supportType]' --output text
done

aws eks describe-cluster-versions lists the support end dates for each version.

RDS instances, with engine version and size:

sh
aws rds describe-db-instances \
  --query 'DBInstances[].{DB:DBInstanceIdentifier,Engine:Engine,EngineVersion:EngineVersion,InstanceClass:DBInstanceClass,MultiAZ:MultiAZ}' \
  --output table

Use describe-db-clusters for Aurora and Multi-AZ DB clusters. Compare the major versions with the dates above. Anything past its end of standard support date is paying, per vCPU, and twice with Multi-AZ.

AWS Health. EKS sends a Health notice about 12 months after a version’s release, which is a couple of months before standard support ends. Make sure those notices reach someone who reads them.

Opting out#

For EKS, set the upgrade policy on the cluster. This works on existing clusters, as long as the version is still in standard support:

hcl
resource "aws_eks_cluster" "this" {
  # ...
  upgrade_policy {
    support_type = "STANDARD"
  }
}

CLI equivalent: aws eks update-cluster-config --name <cluster> --upgrade-policy supportType=STANDARD.

For RDS, use engine_lifecycle_support on aws_db_instance, or on aws_rds_cluster for Aurora and Multi-AZ DB clusters:

hcl
resource "aws_db_instance" "this" {
  # ...
  engine_lifecycle_support = "open-source-rds-extended-support-disabled"
}

RDS lets you change this on existing databases too (aws rds modify-db-instance --engine-lifecycle-support ...), immediately and with no downtime. The AWS provider (v6.73.0 when we checked) only sends the argument when it creates or restores a database. For an existing one, change it with the CLI and set the same value in code so the plan stays clean.

One warning: if you disable enrollment on a database that’s already past end of standard support, RDS upgrades it to the next major version automatically. Don’t do this on production to save money unless you’re ready for that upgrade.

When paying is the better option#

You don’t have to opt out everywhere. Extended support buys time, and sometimes that’s worth the money.

The fee tends to be worth paying when:

  • The upgrade has unknowns. MySQL 8.0 to 8.4 changes authentication defaults, removes deprecated options and needs a new mysql8.4 parameter group. A major PostgreSQL upgrade can change query plans. A few months of fees while you test on a restored snapshot or a Blue/Green deployment costs less than an outage.
  • The instance is small. A 2 vCPU instance in year 1 adds about $146 a month. If the upgrade takes a week of someone’s time plus a maintenance window, paying for a quarter while you plan it is a fair trade.
  • The EKS upgrade spans several versions. EKS upgrades one minor version at a time, and each step means checking deprecated APIs, add-on versions and nodes. Rushing three of those to save $365 a month on one cluster is usually a bad deal.

It stops being worth it when:

  • Instances are large or there are many of them. A 32 vCPU Multi-AZ instance is 64 billable vCPUs. That’s about $4,700 a month in year 1, and double that in year 3.
  • Nobody is planning the upgrade. With no plan and no owner, you pay the fee and still get the forced upgrade at the end.
  • It’s not production. Dev and staging clusters are good candidates for STANDARD. They auto-upgrade, and you find what breaks there before production gets the same upgrade.

The forced upgrade comes either way. Paying only moves it later.

Where we land#

Give extended support an owner. Search the bill for it, list what’s enrolled, and decide for each resource: upgrade now, pay and schedule the upgrade, or set it to standard and accept the auto-upgrade. Put the dates above in your planning.

Extended support is one of the first things we check in a Cost & Performance Review. If you’d like us to go through your bill, get in touch.