Request: starrocks-cluster-sync support for shared-data (S3) target clusters – strip rack/replication properties automatically

Infrastructure and Operations
1

Hello StarRocks team,

We are performing a cross-cluster migration from a shared-nothing StarRocks cluster to a shared-data (S3) cluster, and we are running into a blocking issue when using the official starrocks-cluster-sync tool.

Source cluster:

  • Type: shared-nothing

  • Version: (current production, please note exact build if needed)

Uses rack-aware placement:
“labels.location” = “rack:rack_a,rack:rack_b,rack:rack_c”,
“replication_num” = “3”,
“replicated_storage” = “true”

Target cluster:

    • Type: shared-data (S3)

    • Version: 4.0.0-dccb665

    • 3 Compute Nodes (CNs)

    • Shared storage on S3

    • Does not support ALTER SYSTEM MODIFY BACKEND ... SET ("tag.location" = "...")

Problem:
During migration, the tool fails to create tables on the target cluster with the following error:
java.sql.SQLSyntaxErrorException: Getting analyzing error. Detail message: Cannot find any backend with location: rack:rack_a.

This happens because the target shared-data cluster does not expose rack labels or support rack-based replica placement, so any CREATE TABLE statement containing "labels.location", "replication_num", or "replicated_storage" is rejected.

Even with three CNs, the issue persists because shared-data clusters do not have rack topology metadata.

What we need:
A version of starrocks-cluster-sync that can automatically detect when the target cluster is shared-data (S3) and strip the following properties from all CREATE TABLE statements before executing them:

  • "labels.location"

  • "replication_num"

  • "replicated_storage"

Expected behavior:

  • Tool should create all databases and tables successfully on a shared-data target, even when source DDLs contain rack/replication properties.

  • Tool should continue syncing data and routine jobs as usual after schema creation.

Actual behavior:

  • Tool stops with Cannot find any backend with location: rack:rack_a during CREATE TABLE.

Could you please confirm if such a build already exists (or if there is an internal branch or upcoming release that includes this feature)?
If not, can you provide guidance on obtaining or building a compatible version of starrocks-cluster-sync that supports shared-data targets?

Thank you,
Boyan Bonev