
I Tried Creating a Cross-Continental Multi-Region Cluster (Tokyo and Virginia) with Amazon Aurora DSQL
This page has been translated by machine translation. View original
Introduction
On November 19, 2026, it was announced that Amazon Aurora DSQL multi-Region clusters can now specify Regions on different continents as peers.
Applications on two continents can read from and write to a single strongly consistent database, and even if one Region becomes unavailable, committed writes are preserved and reads and writes can continue in the other Region.
In this article, I create a cluster with Tokyo and N. Virginia as peers and Oregon as the witness, and then read and write through both endpoints.
Test Environment
| Item | Details |
|---|---|
| Peers | ap-northeast-1 (Tokyo), us-east-1 (N. Virginia) |
| witness | us-west-2 (Oregon) |
| AWS CLI | 2.41.2 |
| psql | 18.6 |
Building the Cluster
Specify a different Region for each of the two peers and the witness. Choose from the list of supported Regions.
Specifying Peers and the Witness
In each of the two Regions that will be peers, run create-cluster with the witness specified.
aws dsql create-cluster --region ap-northeast-1 \
--multi-region-properties '{"witnessRegion":"us-west-2"}'
aws dsql create-cluster --region us-east-1 \
--multi-region-properties '{"witnessRegion":"us-west-2"}'
This is the response from Tokyo (N. Virginia had the same shape). The clusters field contains only the ARN of the cluster itself. In the next step, this ARN is specified on the other cluster.
{
...
"status": "PENDING_SETUP",
...
"multiRegionProperties": {
"witnessRegion": "us-west-2",
"clusters": [
"arn:aws:dsql:ap-northeast-1:111122223333:cluster/<tokyo-cluster-id>"
]
},
...
}
Mutual Peering
With update-cluster, specify the other cluster's ARN on each cluster. For witnessRegion, specify the same value as at creation.
aws dsql update-cluster --region ap-northeast-1 --identifier <tokyo-cluster-id> \
--multi-region-properties '{"witnessRegion":"us-west-2","clusters":["arn:aws:dsql:us-east-1:111122223333:cluster/<virginia-cluster-id>"]}'
aws dsql update-cluster --region us-east-1 --identifier <virginia-cluster-id> \
--multi-region-properties '{"witnessRegion":"us-west-2","clusters":["arn:aws:dsql:ap-northeast-1:111122223333:cluster/<tokyo-cluster-id>"]}'
In both responses, the status was UPDATING.
Checking Status and Configuration
Use wait cluster-active to wait until both clusters become ACTIVE.
aws dsql wait cluster-active --region ap-northeast-1 --identifier <tokyo-cluster-id>
aws dsql wait cluster-active --region us-east-1 --identifier <virginia-cluster-id>
It took 8 seconds from the start of create-cluster to the execution of the second update-cluster. The Tokyo wait finished 1 minute 41 seconds after the second update-cluster. When I ran get-cluster after subsequently running the N. Virginia wait, both clusters were ACTIVE (within 2 minutes of peering).
Check the Tokyo configuration with get-cluster.
aws dsql get-cluster --region ap-northeast-1 --identifier <tokyo-cluster-id>
{
...
"status": "ACTIVE",
...
"multiRegionProperties": {
"witnessRegion": "us-west-2",
"clusters": [
"arn:aws:dsql:us-east-1:111122223333:cluster/<virginia-cluster-id>",
"arn:aws:dsql:ap-northeast-1:111122223333:cluster/<tokyo-cluster-id>"
]
},
...
"endpoint": "<tokyo-cluster-id>.dsql.ap-northeast-1.on.aws",
...
}
clusters contains two ARNs. Because endpoints differ by Region, there are two connection targets: Tokyo and N. Virginia.
Reading and Writing Data
How to Connect
Save a script that issues an authentication token and passes it to psql through the environment variable PGPASSWORD as dsql.sh.
#!/usr/bin/env bash
# Issue an authentication token and pass it to psql only via the environment variable PGPASSWORD
# Usage: DSQL_ENDPOINT=<endpoint> DSQL_REGION=<region> ./dsql.sh -c "SQL"
set -euo pipefail
EP="${DSQL_ENDPOINT:?DSQL_ENDPOINT is required}"
REGION="${DSQL_REGION:?DSQL_REGION is required}"
PGPASSWORD="$(aws dsql generate-db-connect-admin-auth-token \
--hostname "$EP" --region "$REGION" --expires-in 3600)" \
PGSSLMODE=require \
psql -h "$EP" -U admin -d postgres "$@"
Grant execute permission, then run it by passing the endpoint and Region of the target.
chmod +x dsql.sh
DSQL_ENDPOINT=<tokyo-cluster-id>.dsql.ap-northeast-1.on.aws DSQL_REGION=ap-northeast-1 ./dsql.sh -c "SQL"
Execution Results
Create a table in Tokyo and write one row.
DSQL_ENDPOINT=<tokyo-cluster-id>.dsql.ap-northeast-1.on.aws DSQL_REGION=ap-northeast-1 \
./dsql.sh -c "CREATE TABLE messages (id uuid PRIMARY KEY DEFAULT gen_random_uuid(), written_from text NOT NULL, body text NOT NULL, created_at timestamptz DEFAULT now())"
CREATE TABLE
DSQL_ENDPOINT=<tokyo-cluster-id>.dsql.ap-northeast-1.on.aws DSQL_REGION=ap-northeast-1 \
./dsql.sh -c "INSERT INTO messages (written_from, body) VALUES ('ap-northeast-1', 'hello from Tokyo') RETURNING written_from, body, created_at"
written_from | body | created_at
----------------+------------------+-------------------------------
ap-northeast-1 | hello from Tokyo | 2026-11-20 01:01:28.279529+00
(1 行)
INSERT 0 1
Run a SELECT in N. Virginia.
DSQL_ENDPOINT=<virginia-cluster-id>.dsql.us-east-1.on.aws DSQL_REGION=us-east-1 \
./dsql.sh -c "SELECT written_from, body, created_at FROM messages ORDER BY created_at"
written_from | body | created_at
----------------+------------------+-------------------------------
ap-northeast-1 | hello from Tokyo | 2026-11-20 01:01:28.279529+00
(1 行)
I did not create the table in N. Virginia, but I could reference the table created in Tokyo, and the row written in Tokyo was returned.
Write one row in N. Virginia.
DSQL_ENDPOINT=<virginia-cluster-id>.dsql.us-east-1.on.aws DSQL_REGION=us-east-1 \
./dsql.sh -c "INSERT INTO messages (written_from, body) VALUES ('us-east-1', 'hello from Virginia') RETURNING written_from, body, created_at"
written_from | body | created_at
--------------+---------------------+------------------------------
us-east-1 | hello from Virginia | 2026-11-20 01:01:36.57591+00
(1 行)
INSERT 0 1
Run a SELECT in Tokyo.
DSQL_ENDPOINT=<tokyo-cluster-id>.dsql.ap-northeast-1.on.aws DSQL_REGION=ap-northeast-1 \
./dsql.sh -c "SELECT written_from, body, created_at FROM messages ORDER BY created_at"
written_from | body | created_at
----------------+---------------------+-------------------------------
ap-northeast-1 | hello from Tokyo | 2026-11-20 01:01:28.279529+00
us-east-1 | hello from Virginia | 2026-11-20 01:01:36.57591+00
(2 行)
In both directions, a SELECT run in the other Region after the INSERT could read the row that had been written.
Summary
I was able to create an Aurora DSQL multi-Region cluster with Tokyo (Asia Pacific) and N. Virginia (North America), which are on different continents, as peers.
I specified Oregon as the witness. The witness Region retains encrypted transaction logs for a limited period, improving multi-Region durability and availability. It has no endpoint, so it is not a connection target for reads or writes.
Previously, the Regions of a multi-Region cluster were limited to the same geographic Region set. For that reason, a peer for Tokyo had to be chosen from Osaka, Seoul, Singapore, and so on. The Regions in the same Region set as Tokyo are as follows.
- Asia Pacific (Hong Kong)
- Asia Pacific (Mumbai)
- Asia Pacific (Osaka)
- Asia Pacific (Seoul)
- Asia Pacific (Singapore)
- Asia Pacific (Tokyo)
With this update, you can create multi-Region clusters with Tokyo and Regions in North America or Europe as peers, making it easier to adopt for DR and for use cases that protect data by keeping it physically far apart.
