I presented at DevelopersIO 2026 Tokyo with the title "Key Points for Using the Snowflake Openflow Connector." #devio2026

This is Kasahara from the Data Business Division.

I gave a presentation titled "Key Points for Using Snowflake Openflow Connector" at "DevelopersIO 2026 Tokyo" held on Saturday, October 24, 2026, so I am sharing the materials here.

Slide Materials

Overview

I have summarized the pitfalls I encountered when using Snowflake Openflow Connector and how to deal with them.

When introducing the Snowflake Openflow Connector, check the following points and proceed accordingly.

Before You Start

• Accept the terms and conditions with ORGADMIN
• Set the DEFAULT_ROLE of the user to something other than ACCOUNTADMIN to access the Runtime
• Check the usage conditions and other constraints of the connector

Network Configuration

• To connect externally from SPCS, configure these three items: "Create Network Rule and EAI", "Grant EAI privileges to the Runtime role", and "Associate EAI with the Runtime"
• If using the Business Critical edition, use PrivateLink for closed network connection / otherwise, since traffic goes through a public network, use NLB and implement IP restriction measures

Preparation on the Source DB Side

• Agree in advance with the DB administrator on licenses, log settings, and users before configuring
• For PostgreSQL: configure Logical Replication settings / Publication settings
• For OracleDB: confirm availability of GoldenGate / XStream, then configure ARCHIVELOG mode / supplemental logging / STREAM_POOL_SIZE, etc.
• Tables to be synchronized also need to be specified on the source DB side

After It Is Running

• Snapshot arrives but CDC does not
  • Check the archive log retention period, etc.
  • If archive logs are not output, changes will not be reflected to Snowflake in the first place
• The order for adding tables to be ingested is "source DB side settings" → "Snowflake side settings"
  • Getting the order wrong can cause data loss without any errors
• If synchronization has stopped even though the status shows "Running", you need to determine whether the issue is on the Oracle side or the Snowflake side
  • For Oracle, check in this order: "capture (Oracle)" → "outbound (Oracle)" → "Connector (Snowflake)"
  • If you still cannot figure it out, try digging deeper into the processes on the Snowflake side
• Set up monitoring and operational rules
  • Basically, do not stop the Connector
  • Before going into production, test and confirm the procedure to verify whether synchronization can catch up without data loss after stopping and restarting

Cleanup

• Be careful not to leave resources that you think you have deleted but are actually still remaining
  • destroy.sh (agent in BYOC) / compute pool / slots / endpoints

If You Are Going to Use Openflow Going Forward

• Use gen2, and build using SQL procedures

References