Skip to main content

Google Cloud SQL for MySQL

This is a change data capture (CDC) connector that captures change events from a MySQL database via the Binary Log.

It is available for use in the Flow web application. For local development or open-source workflows, provides the latest version of the connector as a Docker image. You can also follow the link in your browser to see past image versions.


To use this connector, you'll need a MySQL database setup with the following.

  • binlog_format system variable set to ROW (the default value).
  • Binary log expiration period set to MySQL's default value of 30 days (2592000 seconds) if at all possible.
    • This value may be set lower if necessary, but we strongly discourage going below 7 days as this may increase the likelihood of unrecoverable failures.
  • A watermarks table. The watermarks table is a small "scratch space" to which the connector occasionally writes a small amount of data (a UUID, specifically) to ensure accuracy when backfilling preexisting table contents.
    • The default name is "flow.watermarks", but this can be overridden in config.json.
  • A database user with appropriate permissions:
    • Permission to insert, update, and delete on the watermarks table.
    • Permission to read the tables being captured.
    • Permission to read from information_schema tables, if automatic discovery is used.
  • If the table(s) to be captured include columns of type DATETIME, the time_zone system variable must be set to an IANA zone name or numerical offset or the capture configured with a timezone to use by default.


  1. Allow connections between the database and Estuary Flow. There are two ways to do this: by granting direct access to Flow's IP or by creating an SSH tunnel.

    1. To allow direct access:

    2. To allow secure connections via SSH tunneling:

  2. Set the instance's binlog_expire_logs_seconds flag to 2592000.

  3. Using Google Cloud Shell or your preferred client, create the watermarks table.

CREATE TABLE IF NOT EXISTS flow.watermarks (slot INTEGER PRIMARY KEY, watermark TEXT);
  1. Create the flow_capture user with replication permission, the ability to read all tables, and the ability to read and write the watermarks table.

The SELECT permission can be restricted to just the tables that need to be captured, but automatic discovery requires information_schema access as well.

COMMENT 'User account for Flow MySQL data capture';
GRANT SELECT ON *.* TO 'flow_capture';
GRANT INSERT, UPDATE, DELETE ON flow.watermarks TO 'flow_capture';
  1. In the Cloud Console, note the instance's host under Public IP Address. Its port will always be 3306. Together, you'll use the host:port as the address property when you configure the connector.

Setting the MySQL time zone

MySQL's time_zone server system variable is set to SYSTEM by default.

If you intend to capture tables including columns of the type DATETIME, and time_zone is set to SYSTEM, Flow won't be able to detect the time zone and convert the column to RFC3339 format. To avoid this, you must explicitly set the time zone for your database.

You can:

  • Specify a numerical offset from UTC.

    • For MySQL version 8.0.19 or higher, values from -13:59 to +14:00, inclusive, are permitted.
    • Prior to MySQL 8.0.19, values from -12:59 to +13:00, inclusive, are permitted
  • Specify a named timezone in IANA timezone format.

  • If you're using Amazon Aurora, create or modify the DB cluster parameter group associated with your MySQL database. Set the time_zone parameter to the correct value.

For example, if you're located in New Jersey, USA, you could set time_zone to -05:00 or -04:00, depending on the time of year. Because this region observes daylight savings time, you'd be responsible for changing the offset. Alternatively, you could set time_zone to America/New_York, and time changes would occur automatically.

If using IANA time zones, your database must include time zone tables. Learn more in the MySQL docs.

Capture Timezone Configuration

If you are unable to set the time_zone in the database and need to capture tables with DATETIME columns, the capture can be configured to assume a time zone using the timezone configuration property (see below). The timezone configuration property can be set as a numerical offset or IANA timezone format.

Backfills and performance considerations

When the a MySQL capture is initiated, by default, the connector first backfills, or captures the targeted tables in their current state. It then transitions to capturing change events on an ongoing basis.

This is desirable in most cases, as in ensures that a complete view of your tables is captured into Flow. However, you may find it appropriate to skip the backfill, especially for extremely large tables.

In this case, you may turn of backfilling on a per-table basis. See properties for details.


You configure connectors either in the Flow web app, or by directly editing the catalog specification file. See connectors to learn more about using connectors. The values and specification sample below provide configuration details specific to the MySQL source connector.



/addressServer AddressThe host or host:port at which the database can be reached.stringRequired
/userLogin UserThe database user to authenticate as.stringRequired, "flow_capture"
/passwordLogin PasswordPassword for the specified database user.stringRequired
/timezoneTimezoneTimezone to use when capturing datetime columns. Should normally be left blank to use the database's 'time_zone' system variable. Only required if the 'time_zone' system variable cannot be read and columns with type datetime are being captured. Must be a valid IANA time zone name or +HH:MM offset. Takes precedence over the 'time_zone' system variable if both are set.string
/advanced/watermarks_tableWatermarks Table NameThe name of the table used for watermark writes. Must be fully-qualified in '<schema>.<table>' form.string"flow.watermarks"
/advanced/dbnameDatabase NameThe name of database to connect to. In general this shouldn't matter. The connector can discover and capture from all databases it's authorized to access.string"mysql"
/advanced/node_idNode IDNode ID for the capture. Each node in a replication cluster must have a unique 32-bit ID. The specific value doesn't matter so long as it is unique. If unset or zero the connector will pick a value.integer
/advanced/skip_backfillsSkip BackfillsA comma-separated list of fully-qualified table names which should not be backfilled.string
/advanced/backfill_chunk_sizeBackfill Chunk SizeThe number of rows which should be fetched from the database in a single backfill query.integer131072
/advanced/skip_binlog_retention_checkSkip Binlog Retention Sanity CheckBypasses the 'dangerously short binlog retention' sanity check at startup. Only do this if you understand the danger and have a specific need.boolean


/namespaceNamespaceThe database/schema in which the table resides.stringRequired
/streamStreamName of the table to be captured from the database.stringRequired
/syncModeSync modeConnection method. Always set to incremental.stringRequired

When you configure this connector in the web application, the automatic discovery process sets up a binding for most tables it finds in your database, but there are exceptions.

Tables in the MySQL system schemas information_schema, mysql, performance_schema, and sys will not be discovered. You can add bindings for such tables manually.


A minimal capture definition will look like the following:

address: ""
user: "flow_capture"
password: "secret"
- resource:
namespace: ${TABLE_NAMESPACE}
stream: ${TABLE_NAME}
syncMode: incremental

Your capture definition will likely be more complex, with additional bindings for each table in the source database.

Learn more about capture definitions.

Troubleshooting Capture Errors

The source-mysql connector is designed to halt immediately if something wrong or unexpected happens, instead of continuing on and potentially outputting incorrect data. What follows is a non-exhaustive list of some potential failure modes, and what action should be taken to fix these situations:

Unsupported Operations

If your capture is failing with an "unsupported operation {ALTER,DROP,TRUNCATE,etc} TABLE" error, this indicates that such an operation has taken place impacting a table which is currently being captured.

In the case of DROP TABLE and other destructive operations this is not supported, and can only be resolved by removing the offending table(s) from the capture bindings list, after which you may recreate the capture if desired (causing the latest state of the table to be recaptured in its entirety).

In the case of ALTER TABLE we currently support table alterations to add or drop columns from a table. This error indicates that whatever alteration took place is not currently supported. Practically speaking the immediate resolution is the same as for a DROP or TRUNCATE TABLE, but if you frequently perform schema migrations it may be worth reaching out to see if we can add support for whatever table alteration you just did.

Data Manipulation Queries

If your capture is failing with an "unsupported DML query" error, this means that an INSERT, UPDATE, DELETE or other data manipulation query is present in the MySQL binlog. This should generally not happen if binlog_format = 'ROW' as described in the Prerequisites section.

Resolving this error requires fixing the binlog_format system variable, and then either tearing down and recreating the entire capture so that it restarts at a later point in the binlog, or in the case of an INSERT/DELETE query it may suffice to remove the capture binding for the offending table and then re-add it.

Unhandled Queries

If your capture is failing with an "unhandled query" error, some SQL query is present in the binlog which the connector does not (currently) understand.

In general, this error suggests that the connector should be modified to at least recognize this type of query, and most likely categorize it as either an unsupported DML Query, an unsupported Table Operation, or something that can safely be ignored. Until such a fix is made the capture cannot proceed, and you will need to tear down and recreate the entire capture so that it restarts from a later point in the binlog.

Metadata Errors

If your capture is failing with a "metadata error" then something has gone badly wrong with the capture's tracking of table metadata, such as column names or datatypes.

This should never happen, and most likely means that the MySQL binlog itself is corrupt in some way. If this occurs, it can be resolved by removing the offending table(s) from the capture bindings list and then recreating the capture (generally into a new collection, as this process will cause the table to be re-captured in its entirety).

Insufficient Binlog Retention

If your capture fails with a "binlog retention period is too short" error, it is informing you that the MySQL binlog retention period is set to a dangerously low value, and your capture would risk unrecoverable failure if it were paused or the server became unreachable for a nontrivial amount of time, such that the database expired a binlog segment that the capture was still reading from.

(If this were to happen, then change events would be permanently lost and that particular capture would never be able to make progress without potentially producing incorrect data. Thus the capture would need to be torn down and recreated so that each table could be re-captured in its entirety, starting with a complete backfill of current contents.)

The "binlog retention period is too short" error should normally be fixed by setting binlog_expire_logs_seconds = 2592000 as described in the Prerequisites section (and when running on a managed cloud platform additional steps may be required, refer to the managed cloud setup instructions above). However, advanced users who understand the risks can use the skip_binlog_retention_check configuration option to disable this safety.

Empty Collection Key

Every Flow collection must declare a key which is used to group its documents. When testing your capture, if you encounter an error indicating collection key cannot be empty, you will need to either add a key to the table in your source, or manually edit the generated specification and specify keys for the collection before publishing to the catalog as documented here.