Microsoft SQL Server (MSSQL)
Airbyte’s certified MSSQL connector gives you these features:
- Keep your data fresh through several methods, including Change Data Capture (CDC) using SQL Server’s CDC feature.
- Choose Incremental or Full Refresh sync modes for flexible data delivery to your destination.
- Replicate tables of any size reliably with checkpointing and chunked database reads.
⚠️ Please note the minimum required platform version is v0.58.0 to run source-mssql 4.0.18 and above.
Features
| Feature | Supported | Notes |
|---|---|---|
| Full Refresh Sync | Yes | |
| Incremental Sync - Append | Yes | |
| Replicate Incremental Deletes | Yes | |
| CDC (Change Data Capture) | Yes | |
| SSL Support | Yes | |
| SSH Tunnel Connection | Yes | |
| Microsoft Entra ID Auth | Yes | Service principal |
| Namespaces | Yes | Enabled by default |
The MSSQL source doesn’t alter your database schema. However, the connected destination may alter it. See the destination’s documentation for details.
Get started
Requirements
- MSSQL Server
Azure SQL Database,Azure Synapse Analytics,Azure SQL Managed Instance,SQL Server 2022,SQL Server 2019,SQL Server 2017,SQL Server 2016,SQL Server 2014,SQL Server 2012,PDW 2008R2 AU34. - Create a dedicated, read-only Airbyte user with access to every table you need to replicate.
- To use CDC, review the relevant section below for further setup requirements.
1. Make your database accessible to Airbyte
Access depends on your network setup. Use the UI’s check connection tool to confirm Airbyte can connect to your MSSQL instance.
For Airbyte Cloud, your firewall or security group must allow the Airbyte Cloud IP addresses. The connection must also be encrypted. See Airbyte Cloud below.
2. Create a dedicated, read-only user for the relevant tables (recommended but optional)
This optional step gives you better permission control and auditing. Alternatively, you can connect Airbyte using an existing database user.
-
Create an Airbyte login and database user, then add the user to the db_datareader role. Membership in
db_datareadergrantsSELECTon every current and future database table:USE {database name}; CREATE LOGIN {user name} WITH PASSWORD = '{password}'; CREATE USER {user name} FOR LOGIN {user name}; ALTER ROLE db_datareader ADD MEMBER {user name};ALTER ROLE ... ADD MEMBERreplaces the deprecatedsp_addrolememberstored procedure, which Microsoft recommends avoiding in new work. -
To limit access to specific schemas, skip the
db_datareaderrole. Instead, grantSELECTon each schema you want to replicate. Run this command for every schema:USE {database name}; GRANT SELECT ON SCHEMA :: {schema name} TO {user name};
Use this username and password when configuring the MSSQL source in Airbyte. For CDC, the user also needs permissions described in Setting up CDC for MSSQL.
3. Your database user is ready for Airbyte
Airbyte Cloud
Airbyte Cloud only supports secured connections to your MSSQL instance in the source configuration. It never connects to a database without encryption.
The connection must use TLS/SSL or an SSH tunnel. The tunnel encrypts the connection between Airbyte and your bastion host. In practice:
| Setup | Supported on Airbyte Cloud |
|---|---|
Encryption set to Encrypted (trust server certificate) or Encrypted (verify certificate) | Yes, with or without an SSH tunnel. |
Encryption set to Unencrypted with an SSH tunnel | Yes. The SSH tunnel provides the encryption. |
Encryption set to Unencrypted without an SSH tunnel | No. The connection test fails with Connection from Airbyte Cloud requires SSL encryption or an SSH tunnel. |
Database on localhost or on a private network (not reachable from the public internet) | Only through an SSH tunnel to a bastion host that can reach the database, which also provides the encryption Airbyte Cloud requires. With No Tunnel, localhost, 127.0.0.1 or 0.0.0.0 fails the connection test with Airbyte Cloud cannot reach a database on your own machine. Use a hostname reachable from the internet, or configure an SSH tunnel. |
A URL, credentials or host:port pasted into Host | No. The connection test fails with Enter only the hostname in Host; put the port in the Port field. Enter only the hostname or IP address in Host and the port in Port. |
On Airbyte Cloud, Encryption defaults to Encrypted (trust server certificate).
See Connection to MSSQL via an SSH Tunnel to set up the tunnel. If your network restricts inbound connections, see IP allow list.
Microsoft Entra ID authentication
This connector supports Microsoft Entra ID authentication, formerly Azure Active Directory, through a service principal. It’s an alternative to SQL Server username and password authentication.
This authentication mode is recommended for Azure SQL Database and Azure SQL Managed Instance.
Prerequisites
- An Azure SQL Database, Azure SQL Managed Instance or SQL Server instance configured for Microsoft Entra authentication. For setup instructions, see Configure and manage Microsoft Entra authentication with Azure SQL.
- A Microsoft Entra ID app registration (service principal) with a client secret.
- A database user created for the service principal, with
SELECTaccess to the tables you want to replicate. For instructions, see Create Microsoft Entra users using service principals. - For CDC, the service principal user also needs the CDC permissions described in Setting up CDC for MSSQL.
Configuration
Complete these fields under Microsoft Entra ID in the source configuration form:
| Field | Description |
|---|---|
| Entra ID Client ID | The application (client) ID of the service principal. |
| Entra ID Client Secret | The client secret generated for the service principal. |
When you provide both fields, the connector authenticates in ActiveDirectoryServicePrincipal mode through the Microsoft JDBC driver. It ignores Username and Password.
If either Entra ID field is empty, the connector uses username and password authentication.
:::note
Entra ID authentication requires an encrypted connection. Set Encryption to Encrypted (trust server certificate) or Encrypted (verify certificate).
The configuration check fails if you set Entra ID fields while encryption is disabled. :::
MSSQL replication modes
Incremental syncs
MSSQL datetime2(7) stores timestamps with up to seven decimal places. Most destinations don’t support this precision, so Airbyte truncates timestamps to six decimal places, or microseconds.
The saved cursor can therefore fall just behind your table’s newest rows. To prevent skipped rows, each incremental sync reads everything above the saved cursor.
It then saves the new maximum value as the cursor for the next sync.
:::note
Each sync collects everything newer than the last saved point, so you may see duplicate rows. Where supported, Incremental - Append + Deduped is recommended for these syncs.
For more information, see the Sync Mode page.
:::
Change Data Capture (CDC)
Airbyte uses
SQL Server’s change data capture feature
with transaction logs to capture row-level INSERT, UPDATE and DELETE operations on CDC-enabled tables.
Additional setup requires at least db_owner permissions on each database you intend to sync. See below for details.
Read CDC docs for an overview of Airbyte’s approach to CDC.
Should you use CDC for MSSQL?
- Use CDC if you need deletion records and can accept the limitations below.
- For a small dataset or destination snapshot, consider Full Refresh replication instead of CDC.
- If CDC’s limitations prevent its use, consider non-CDC incremental replication to maintain a snapshot. Occasionally reset the data and sync again.
- If your table has a primary key but no suitable incremental cursor, such as
updated_at, CDC supports incremental syncing.
CDC limitations
- Read CDC docs for limitations affecting every database that uses CDC replication.
hierarchyidandsql_varianttypes aren’t processed with the CDC migration type because Debezium doesn’t support them. For details, see this ticket.- CDC is only available for SQL Server 2016 Service Pack 1 (SP1) and later.
- db_owner (or higher) permissions are required to perform the necessary setup for CDC.
- On Linux, CDC isn’t supported before SQL Server 2017 CU18. SQL Server 2019 is supported.
- You can’t enable change data capture on tables with a clustered columnstore index. You can enable it with a non-clustered columnstore index.
- SQL Server CDC only processes changes in user-created tables. You can’t enable CDC on the SQL Server master database.
- The
ALTER TABLE…SWITCH TO…PARTITION… statement doesn’t support variables with partition switching on CDC-enabled databases or tables. - CDC incremental syncing only supports tables with at least one primary key. CDC can replicate tables without primary keys only in Full Refresh mode. For more information about CDC limitations, see CDC Limitations doc.
- Airbyte’s CDC implementation provides at-least-once delivery for every change record.
- Read more about CDC limitations in Microsoft docs.
Set up CDC for MSSQL
1. Enable CDC for the database and tables
MS SQL Server provides built-in stored procedures for enabling CDC.
-
First, a SQL Server administrator with db_owner or sysadmin privileges must enable CDC at the database level.
USE {database name} GO EXEC sys.sp_cdc_enable_db GO -
The administrator must then enable CDC for every table you want to capture. Here’s an example:
USE {database name} GO EXEC sys.sp_cdc_enable_table @source_schema = N'{schema name}', @source_name = N'{table name}', @role_name = N'{role name}', [1] @filegroup_name = N'{filegroup name}', [2] @supports_net_changes = 0 [3] GO- [1] Specifies a role that receives
SELECTpermission on the source table’s captured columns. Set a value so you can use this role in the next step. Alternatively, set @rolename toNULL. This restricts access to _sysadmin and db_owner. Ensure the Airbyte source credentials match this role, so Airbyte can access the cdc tables. - [2] Specifies the filegroup where SQL Server stores the change table. A separate CDC filegroup is recommended. You can omit this parameter to use the default filegroup.
- [3] When set to 0, SQL Server only generates support functions for querying all changes. When set to 1, it also generates functions for querying net changes. If supports_net_changes is 1, specify index_name or ensure the source table has a defined primary key.
- [1] Specifies a role that receives
-
For more information about the parameters, see Microsoft doc page for this stored procedure.
-
If you need to enable CDC on many tables, use this script to avoid running this query separately for every table.
For more detail, see Microsoft docs on enabling and disabling CDC.
:::note Google Cloud SQL for SQL Server
On Google Cloud SQL for SQL Server, Google doesn’t grant customers the sysadmin server role. Therefore, you can’t run sys.sp_cdc_enable_db to enable database-level CDC.
Instead, use Cloud SQL’s dedicated stored procedure. It enables CDC without sysadmin:
EXEC msdb.dbo.gcloudsql_cdc_enable_db 'YOUR_DATABASE_NAME'
To disable database-level CDC later, use the corresponding
EXEC msdb.dbo.gcloudsql_cdc_disable_db 'YOUR_DATABASE_NAME' procedure.
Only database-level enablement differs on Cloud SQL. Enabling CDC on individual tables still uses the standard sys.sp_cdc_enable_table procedure.
The snapshot isolation and user permission steps below remain unchanged. For Google’s complete procedure, see Configure CDC for a Cloud SQL for SQL Server source.
:::
2. Enable snapshot isolation
-
During the first CDC sync, Airbyte takes an initial, consistent database snapshot. Airbyte uses snapshot isolation to avoid table locks. This approach allows other database clients to write simultaneously. Enable it on the database as follows:
ALTER DATABASE {database name} SET ALLOW_SNAPSHOT_ISOLATION ON;
3. Create a user and grant permissions
-
Create a new Airbyte user with the relevant CDC access instead of using sysadmin or db_owner credentials. First, create the login and user. Then add the user to the db_datareader role:
USE {database name}; CREATE LOGIN {user name} WITH PASSWORD = '{password}'; CREATE USER {user name} FOR LOGIN {user name}; ALTER ROLE db_datareader ADD MEMBER {user name};-
Add the user to the role you specified when enabling cdc on the table(s):
ALTER ROLE {role name} ADD MEMBER {user name}; -
ALTER ROLE ... ADD MEMBERreplaces the deprecatedsp_addrolememberstored procedure, which Microsoft recommends avoiding in new work. -
This access should be sufficient. If you encounter problems, also grant the user direct
SELECTaccess to the cdc schema:USE {database name}; GRANT SELECT ON SCHEMA :: [cdc] TO {user name}; -
Where feasible, grant this user ‘VIEW SERVER STATE’ permissions. Airbyte can then check whether the SQL Server Agent is running. This is preferred because syncs fail if the Agent isn’t updating CDC tables in the source database.
USE master; GRANT VIEW SERVER STATE TO {user name};
-
4. Extend CDC data retention
-
By default, SQL Server retains only three days of data in change tables. Increase retention unless you run very frequent syncs. Longer retention provides time to restart from the last incremental sync point after a failure or pause. Airbyte recommends retaining at least 7 days of CDC data.
-
Change these settings with the sys.sp_cdc_change_job stored procedure, as shown below:
-- Airbyte recommends at least 10080 minutes (7 days) as the retention period EXEC sp_cdc_change_job @job_type='cleanup', @retention = 10080 -
After this change, restart the cleanup job:
EXEC sys.sp_cdc_stop_job @job_type = 'cleanup';
EXEC sys.sp_cdc_start_job @job_type = 'cleanup';
- If you use Transaction Replication, change retention with these scripts:
EXEC sp_changedistributiondb
@database = 'distribution',
@property = 'max_distretention',
@value = 10080 -- 10080 minutes (7 days)
EXEC sp_changedistributiondb
@database = 'distribution',
@property = 'history_retention',
@value = 10080 -- 10080 minutes (7 days)
USE [msdb]
GO
EXEC msdb.dbo.sp_update_jobstep @job_name=N'Distribution clean up: distribution', @step_id=1 ,
@command=N'EXEC dbo.sp_MSdistribution_cleanup @min_distretention = 0, @max_distretention = 10800'
GO
5. Ensure the SQL Server Agent is running
- MSSQL uses the SQL Server Agent to run the jobs necessary for CDC. The Agent must run for CDC to work correctly. Check the SQL Server Agent’s status as follows:
EXEC xp_servicecontrol 'QueryState', N'SQLServerAGENT';
- If you see a status other than ‘Running.’, follow Microsoft docs to start the service.
Connect to MSSQL through an SSH tunnel
Airbyte can connect to an MSSQL instance through an SSH tunnel. Use a tunnel when direct database access is unavailable or violates security policy.
For example, the database may not have a public IP address.
An SSH tunnel connects Airbyte through an intermediate server, also called a bastion server. The bastion has direct database access.
Airbyte connects to the bastion, which then connects directly to the database server. Set the source’s Host and Port to the database address visible from the bastion.
For example, use localhost and 1433 when SQL Server runs on the bastion.
:::note Airbyte Cloud Airbyte Cloud never connects to a database without encryption. An SSH tunnel meets this requirement by encrypting the connection to the bastion.
On Airbyte Cloud, a tunnel is the only way to use the Unencrypted encryption method. It’s also required for databases on localhost or private networks.
Without a tunnel, Airbyte Cloud requires an encrypted SSL method. See Airbyte Cloud. :::
This feature needs additional source configuration. The following steps explain each configuration field.
-
Configure every source field as usual, except
SSH Tunnel Method. -
SSH Tunnel Methoddefaults toNo Tunnel(a direct connection). To use anSSH tunnel, choose
SSH Key AuthenticationorPassword Authentication.-
Choose
Key Authenticationwhen using an RSA private key as your secret forestablishing the SSH tunnel. See below for instructions on generating this key.
-
Choose
Password Authenticationwhen using a password as your secret for establishingthe SSH tunnel.
-
-
SSH Tunnel Jump Server Hostidentifies the intermediate (bastion) server that Airbyte connects to. Entera hostname or IP address.
-
SSH Connection Portis the bastion server port used for the SSH connection. The default port forSSH connections is
22. Use this default unless you’ve explicitly changed the port. -
SSH Login Usernameis the username Airbyte uses to connect to the bastion server. It isn’t theMSSQL username.
-
If you’re using
Password Authentication, setSSH Login Usernameto theprevious step’s user password. If you’re using
SSH Key Authentication, leave thisblank. This isn’t the MSSQL password. It’s the password for the operating system user that Airbyte uses
to run commands on the bastion.
-
If you’re using
SSH Key Authentication, setSSH Private Keyto the RSAprivate key used to create the SSH connection. Include the key file’s complete contents,
starting with
-----BEGIN RSA PRIVATE KEY-----and endingwith
-----END RSA PRIVATE KEY-----.
Generate an SSH key pair
The connector requires an RSA key in PEM format. Generate it as follows:
ssh-keygen -t rsa -m PEM -f myuser_rsa
This creates the private key in PEM format. The public key remains in the standard format used by the bastion host’s authorized_keys file.
Add the public key to the bastion host for the user you want Airbyte to use. Paste the private key into Airbyte’s connector configuration screen.
The connector can then log in to the bastion.
Data type mapping
MSSQL data types map to the following types during synchronisation. See the test value examples here.
If your required data type is missing or causes problems, you can add a new test.
| MSSQL Type | Resulting Type | Notes |
|---|---|---|
bigint | integer | |
binary | binary | |
bit | boolean | |
char | string | |
date | date | |
datetime | timestamp | |
datetime2 | timestamp | |
datetimeoffset | timestamp with timezone | |
decimal | number / integer | maps to integer when the column scale is 0 |
int | integer | |
float | number | |
geography | string | |
geometry | string | |
money | number | |
numeric | number / integer | maps to integer when the column scale is 0 |
ntext | string | |
nvarchar | string | |
nvarchar(max) | string | |
real | number | |
smalldatetime | timestamp | |
smallint | integer | |
smallmoney | number | |
sql_variant | string | |
uniqueidentifier | string | |
text | string | |
time | time | |
tinyint | integer | |
varbinary | binary | |
varchar | string | |
varchar(max) COLLATE Latin1_General_100_CI_AI_SC_UTF8 | string | |
hierarchyid | string | Non-CDC only |
xml | string |
If a type isn’t listed, assume it’s coerced into a string. Feedback on preferred mappings is welcome.
User-defined alias types
SQL Server supports alias types based on built-in types, such as CREATE TYPE alias_timestamp FROM datetime. The connector doesn’t map alias types by name.
Instead, it uses the column’s JDBC type reported by the SQL Server driver. This type comes from the alias’s base type.
The connector maps aliases over these base types like the base types themselves, including during CDC syncs:
| Alias base type | JDBC type | Resulting Type |
|---|---|---|
date | DATE | date |
time | TIME | time |
datetime | TIMESTAMP | timestamp |
datetime2 | TIMESTAMP | timestamp |
smalldatetime | TIMESTAMP | timestamp |
Aliases over unlisted base types may cause mapping issues. Their JDBC types may not map like the built-in types.
Use built-in SQL Server types instead of alias types for columns you sync with Airbyte.
Upgrade to version 4.3.0 or later
Version 4.3.0 moves the connector from the legacy CDK to the new CDK architecture. This migration includes:
- Automatic State Migration: The connector automatically migrates legacy version 2 state formats to the new version 3 format. This includes:
OrderedColumnLoadStatus(primary key-based initial sync) → version 3primary_keystate typeCursorBasedStatus(cursor-based incremental) → version 3cursor_basedstate type
- Backward Compatibility: Existing connections continue working without manual intervention.
Upgrade from 0.4.17 or earlier to 0.4.18 or later
Microsoft SQL Source connector versions 0.4.17 and 0.4.18 have a backwards-incompatible specification change.
The change updates the replication_method configuration parameter to object from string.
In Microsoft SQL source connector version 0.4.17 and earlier, the replication_method configuration parameter was saved in the configuration database as follows:
"replication_method": "STANDARD"
From version 0.4.18, the replication_method configuration parameter is saved as follows:
"replication_method": {
"method": "STANDARD"
}
After upgrading from Microsoft SQL Source connector version 0.4.17 or earlier, fix the source configurations in Airbyte’s actor table.
You must run two SQL queries. Follow Airbyte documentation to run SQL queries on the Airbyte database.
For Microsoft SQL Source connections using the Standard replication method, run this SQL:
update public.actor set configuration =jsonb_set(configuration, '{replication_method}', '{"method": "STANDARD"}', true)
WHERE actor_definition_id ='b5ea17b1-f170-46dc-bc31-cc744ca984c1' AND (configuration->>'replication_method' = 'STANDARD');
For Microsoft SQL Source connections using the Logical Replication (CDC) method, run this SQL:
update public.actor set configuration =jsonb_set(configuration, '{replication_method}', '{"method": "CDC"}', true)
WHERE actor_definition_id ='b5ea17b1-f170-46dc-bc31-cc744ca984c1' AND (configuration->>'replication_method' = 'CDC');
IP allow list
If you use Airbyte Cloud and restrict access to specific IPs, add the Airbyte Cloud IP addresses to your allow list.
If Airbyte can’t open a TCP connection to your server, connection tests and syncs fail with:
SQL Server Connection Error: Airbyte could not reach SQL Server at the configured host and port.
Check the host and port, that the instance accepts remote TCP connections, and that your firewall
or security group allows Airbyte's IP addresses.
Possible causes include an incorrect host or port, an unresolved hostname or an instance that rejects remote TCP connections.
A firewall or security group may also block Airbyte Cloud IP addresses. Airbyte Cloud can’t reach databases on localhost or private networks without an SSH tunnel.
Before connecting, Airbyte Cloud rejects localhost with No Tunnel. It also rejects Airbyte Cloud cannot reach a database on your own machine. Use a hostname reachable from the internet, or configure an SSH tunnel. A URL or host:port pasted into Host with Enter only the hostname in Host; put the port in the Port field..
Changelog
Expand to review
| Version | Date | Pull Request | Subject |
|---|---|---|---|
| 5.1.5 | 2026-11-09 | 89073 | Fix CDC sync failure for columns using user-defined alias types over date, time, datetime, datetime2, or smalldatetime. |
| 5.1.4 | 2026-11-02 | 73349 | Escape spaces and $ in Debezium message keys, and quote schema names in the Debezium schema include list. |
| 5.1.3 | 2026-10-21 | 89634 | Upgrade to Bulk CDK 1.2.3: fix critical CVEs in Apache MINA (mina-core 2.0.31), MINA SSHD 2.20.0 and BouncyCastle 1.86; Kafka 4.1.2, Jetty 12.0.40, plexus-utils 3.6.2. |
| 5.1.2 | 2026-10-19 | 85353 | Cold-start a stream when its saved state is an empty object or an unrecognized legacy format instead of skipping it. |
| 5.1.1 | 2026-10-12 | 88500 | Replace the sync-oriented retry wording of unreachable-host errors with guidance on the host, port, firewall, and Airbyte Cloud IP allow list; note that Unencrypted needs an SSH tunnel on Airbyte Cloud. Explain the Discovered zero tables. connection-test failure: say that no table is visible to the login and how to grant SELECT. On Airbyte Cloud, default the Encryption option to Encrypted (trust server certificate). On Airbyte Cloud, reject localhost without an SSH tunnel and a URL or host:port pasted into Host with an explicit message before the connection is attempted. |
| 5.1.0 | 2026-10-12 | 86318 | Add Debezium’s max_iteration_transactions (max.iteration.transactions) as an advanced CDC option, defaulting to 500. |
| 5.0.2 | 2026-10-09 | 88429 | Upgrade to Bulk CDK 1.1.14. |
| 5.0.1 | 2026-09-09 | 84356 | Change Debezium data.query.mode from direct to function to prevent syncs from hanging on large tables. |
| 5.0.0 | 2026-06-23 | 77672 | Map DECIMAL/NUMERIC columns with scale 0 to Airbyte integer instead of number so destinations preserve integral semantics. |
| 4.4.12 | 2026-06-17 | 80156 | Log a message when a DECIMAL/NUMERIC column with scale 0 is discovered, ahead of an upcoming number -> integer remapping. No functional change. |
| 4.4.11 | 2026-06-11 | 79128 | Fix incremental sync failure when the saved state has a null cursor (table was empty on prior CDK version). |
| 4.4.10 | 2026-06-10 | 79149 | Update cursor-based incremental query to prevent missing rows with high-precision datetime cursors. |
| 4.4.9 | 2026-06-02 | 77998 | Validate CDC access per configured stream to prevent zero-LSN errors caused by missing permissions. |
| 4.4.8 | 2026-05-26 | 78415 | Classify Azure read-replica error 3947 as transient so affected syncs retry instead of failing. |
| 4.4.7 | 2026-05-12 | 78033 | Re-release the Java connector base image revert after the 4.4.6 publish failure. |
| 4.4.6 | 2026-05-07 | 77856 | Revert the Java connector base image to resolve connection issues and remove registry rollback overrides. |
| 4.4.5 | 2026-05-07 | 77843 | Roll back source mssql to 4.4.3 to investigate a potential connection issue. |
| 4.4.4 | 2026-05-07 | 77665 | Fix sampling sync failure on empty tables in Full Refresh mode (NULL upper bound). |
| 4.4.3 | 2026-05-05 | 77786 | Make the hidden additional properties fields in spec optional. No functional change. |
| 4.4.2 | 2026-04-29 | 77036 | Fix TABLESAMPLE failure on views and tables without an ordered column in cursor-incremental syncs. |
| 4.4.1 | 2026-04-23 | 76857 | Fix Invalid column name error when sampling system-versioned temporal tables that have HIDDEN period columns. |
| 4.4.0 | 2026-04-23 | 76143 | Add Microsoft Entra ID service principal authentication for both JDBC and CDC paths. |
| 4.3.6 | 2026-04-20 | 74729 | Fix snapshot partitions restarting from the beginning of the table instead of resuming from the last checkpoint. |
| 4.3.5 | 2026-02-23 | 73606 | Fix CDC cursor overflow. |
| 4.3.4 | 2026-02-17 | 72935 | Update LSN validation to correctly detect when saved offset has been truncated. |
| 4.3.3 | 2026-02-03 | 71821 | Require a manual refresh when schema history is missing, bump CDK version. |
| 4.3.2 | 2025-12-11 | 70836 | Add Azure SQL Database compatibility for SQL Server Agent check |
| 4.3.1 | 2025-12-09 | 70823 | Bump up connector version number to release the connector |
| 4.3.0-rc.10 | 2025-12-05 | 70306 | Update CDK version to include fix for Debezium closing record race condition |
| 4.3.0-rc.9 | 2025-12-01 | 70280 | Fix connector concurrency default value |
| 4.3.0-rc.8 | 2025-11-14 | 69754 | Update to latest CDK version |
| 4.3.0-rc.7 | 2025-11-14 | 69327 | Change default schema to discover all schema instead of “dbo” |
| 4.3.0-rc.6 | 2025-11-12 | 69311 | Fix NPE in state migration when incremental_state is null; Reduce resourceAcquisitionHeartbeat to 100ms; Upgrade CDK version |
| 4.3.0-rc.5 | 2025-11-08 | 69248 | Add SQL Server identifier quoting to fix reserved keyword issues |
| 4.3.0-rc.4 | 2025-11-05 | 69194 | Fix composite primary key discovery to show all PK columns; separate PK discovery from sync strategy |
| 4.3.0-rc.3 | 2025-10-31 | 69097 | Fix connector state value type from string to json object |
| 4.3.0-rc.2 | 2025-10-29 | 69093 | Fix state parsing error for integer and number |
| 4.3.0-rc.1 | 2025-10-28 | 63731 | Migrate source mssql from old CDK to new CDK |
| 4.2.6 | 2025-10-14 | 68091 | Fix CDC table filtering to prevent unintended tables from being synced when using schema-level CDC |
| 4.2.5 | 2025-08-13 | 64905 | bumping up java cdk version for mssql |
| 4.2.4 | 2025-07-03 | 62491 | Improve Debezium performance by configuring the poll interval parameter. |
| 4.2.3 | 2025-07-01 | 62052 | Revert change to CDK interface signature. |
| 4.2.2 | 2025-06-25 | 61729 | Support the use of logical primary keys for CDC. |
| 4.2.1 | 2025-06-23 | 62015 | Fix previous merge. Improve cutoff date handling |
| 4.2.0 | 2025-06-19 | 61685 | Add an option to exclude today’s data from cursor based incremental syncs when using temporal cursor (datetime). |
| 4.1.29 | 2025-06-03 | 61320 | Add error handling for connection issues and adopt the latest CDK version. |
| 4.1.28 | 2025-05-15 | 60311 | Migrate to new gradle flow. |
| 4.1.27 | 2025-04-28 | 59124 | Fix _ab_cdc_event_serial_no datatype in addMetaDataToRowsFetchedOutsideDebezium method |
| 4.1.26 | 2025-03-27 | 56401 | Fix non-unique value clustered index issue |
| 4.1.25 | 2025-03-20 | 55878 | Integrate with the latest mssql jdbc driver to solve a pre-gregorian date offset bug |
| 4.1.24 | 2025-03-11 | 55709 | Filter unwanted tables in discover to prevent null table issues |
| 4.1.23 | 2025-03-06 | 55234 | Update base image version for certified DB source connectors |
| 4.1.22 | 2025-02-10 | 53217 | Default to PK when the clustered index is composite. |
| 4.1.21 | 2025-02-21 | 54189 | Print state data only in the debugging log. |
| 4.1.20 | 2025-01-26 | 52556 | Improve tables discovery during read. |
| 4.1.19 | 2025-01-16 | 51596 | Bump driver versions to latest (jdbc, debezium, cdk) |
| 4.1.18 | 2025-01-06 | 50943 | Use airbyte/java-connector-base:2.0.0. This makes the image rootless. The connector will be incompatible with Airbyte < 0.64. |
| 4.1.17 | 2024-12-17 | 49840 | Use a base image: airbyte/java-connector-base:1.0.0 |
| 4.1.16 | 2024-11-13 | 48484 | Enhanced error handling for MSSQL to improve system error detection and response. |
| 4.1.15 | 2024-10-05 | 46515 | Improving discovery of large SQL server database. |
| 4.1.14 | 2024-09-17 | 45639 | Adopt latest CDK to use the latest apache sshd mina to handle tcpkeepalive requests. |
| 4.1.13 | 2024-09-05 | 45181 | Fix incorrect categorizing resumable/nonresumable full refresh streams. |
| 4.1.12 | 2024-09-10 | 45368 | Remove excessive debezium logging. |
| 4.1.11 | 2024-09-04 | 45142 | Fix incorrect datetimeoffset format in cursor state. |
| 4.1.10 | 2024-08-27 | 44759 | Improve null safety in parsing debezium change events. |
| 4.1.9 | 2024-08-27 | 44841 | Adopt latest CDK. |
| 4.1.8 | 2024-08-08 | 43410 | Adopt latest CDK. |
| 4.1.7 | 2024-08-06 | 42869 | Adopt latest CDK. |
| 4.1.6 | 2024-07-30 | 42550 | Correctly report stream states. |
| 4.1.5 | 2024-07-29 | 42852 | Bump CDK version to latest to use new bug fixes on error translation. |
| 4.1.4 | 2024-07-23 | 42421 | Remove final transient error emitter iterators. |
| 4.1.3 | 2024-07-22 | 42411 | Hide the “initial load timeout in hours” field by default in UI |
| 4.1.2 | 2024-07-22 | 42024 | Fix a NPE bug on resuming from a failed attempt. |
| 4.1.1 | 2024-07-19 | 42122 | Improve wass error message + logging. |
| 4.1.0 | 2024-07-17 | 42078 | WASS analytics + bug fixes. |
| 4.0.36 | 2024-07-17 | 41648 | Implement WASS. |
| 4.0.35 | 2024-07-05 | 40570 | Bump debezium-connector-sqlserver from 2.6.1.Final to 2.6.2.Final. |
| 4.0.34 | 2024-07-01 | 40516 | Remove dbz hearbeat. |
| 4.0.33 | 2024-06-30 | 40638 | Fix a bug that could slow down an initial load of a large table using a different clustered index from the primary key. |
| 4.0.32 | 2024-06-26 | 40558 | Handle DatetimeOffset correctly. |
| 4.0.31 | 2024-06-14 | 39419 | Handle DatetimeOffset correctly. |
| 4.0.30 | 2024-06-14 | 39349 | Full refresh stream sending internal count metadata. |
| 4.0.29 | 2024-06-14 | 39506 | Adopt latest CDK. |
| 4.0.28 | 2024-06-08 | 39342 | Fix custom conversion in CDC for datetimeoffset type. |
| 4.0.27 | 2024-05-29 | 38584 | Set is_resumable flag in discover. |
| 4.0.26 | 2024-05-16 | 38292 | Improve cursor value query to return only one row |
| 4.0.25 | 2024-05-29 | 38775 | Publish CDK |
| 4.0.24 | 2024-05-23 | 38640 | Sync sending trace status messages indicating progress. |
| 4.0.23 | 2024-05-15 | 38208 | disable counts in full refresh stream in state message. |
| 4.0.22 | 2024-05-14 | 38196 | Bump jdbc driver version to 12.6.1.jre11 |
| 4.0.21 | 2024-05-07 | 38054 | Resumeable refresh should run only if there is source defined pk. |
| 4.0.20 | 2024-05-07 | 38042 | Bump debezium version to latest. |
| 4.0.19 | 2024-05-07 | 38029 | Fix previous release. |
| 4.0.18 | 2024-04-30 | 37451 | Resumable full refresh read of tables. |
| 4.0.17 | 2024-05-02 | 37781 | Adopt latest CDK. |
| 4.0.16 | 2024-05-01 | 37742 | Adopt latest CDK. Remove Debezium retries. |
| 4.0.15 | 2024-04-22 | 37541 | Adopt latest CDK. reduce excessive logs. |
| 4.0.14 | 2024-04-22 | 37476 | Adopt latest CDK. |
| 4.0.13 | 2024-04-16 | 37111 | Populate null values in record message. |
| 4.0.12 | 2024-04-15 | 37326 | Allow up to 60 minutes of wait for the an initial CDC record. |
| 4.0.11 | 2024-04-15 | 37325 | Populate airbyte_meta.changes + error handling. |
| 4.0.10 | 2024-04-15 | 37110 | Internal cleanup. |
| 4.0.9 | 2024-04-10 | 36919 | Fix a bug in conversion of null values. |
| 4.0.8 | 2024-04-05 | 36872 | Update to connector’s metadat definition. |
| 4.0.7 | 2024-04-03 | 36772 | Adopt latest CDK. |
| 4.0.6 | 2024-03-25 | 36333 | Deprecate Dbz state iterator. |
| 4.0.5 | 2024-03-21 | 36364 | Allow up to 40 minutes of wait for the an initial CDC record. |
| 4.0.4 | 2024-03-20 | 36325 | [Refactor] : Remove mssql initial source operations . |
| 4.0.3 | 2024-03-19 | 36263 | Fix a failure seen in CDC with tables containing default values. |
| 4.0.2 | 2024-03-06 | 35792 | Initial sync will now send record count in state message. |
| 4.0.1 | 2024-03-12 | 36011 | Read correctly null values of columns with default value in CDC. |
| 4.0.0 | 2024-03-06 | 35873 | Terabyte-sized tables support, reliability improvements, bug fixes. |
| 3.7.7 | 2024-03-06 | 35816 | Fix query that was failing on a case sensitive server. |
| 3.7.6 | 2024-03-04 | 35721 | Fix tests |
| 3.7.5 | 2024-02-29 | 35739 | Allow configuring the queue size used for cdc events. |
| 3.7.4 | 2024-02-26 | 35566 | Add config to throw an error on invalid CDC position. |
| 3.7.3 | 2024-02-23 | 35596 | Fix a logger issue |
| 3.7.2 | 2024-02-21 | 35368 | Change query syntax to make it compatible with Azure SQL Managed Instance. |
| 3.7.1 | 2024-02-20 | 35405 | Change query syntax to make it compatible with Azure Synapse. |
| 3.7.0 | 2024-01-30 | 33311 | Source mssql with checkpointing initial sync. |
| 3.6.1 | 2024-01-26 | 34573 | Adopt CDK v0.16.0. |
| 3.6.0 | 2024-01-10 | 33700 | Remove CDC config options for data_to_sync and snapshot isolation. |
| 3.5.1 | 2024-01-05 | 33510 | Test-only changes. |
| 3.5.0 | 2023-12-19 | 33071 | Fix SSL configuration parameters |
| 3.4.1 | 2024-01-02 | 33755 | Encode binary to base64 format |
| 3.4.0 | 2023-12-19 | 33481 | Remove LEGACY state flag |
| 3.3.2 | 2023-12-14 | 33505 | Using the released CDK. |
| 3.3.1 | 2023-12-12 | 33225 | extracting MsSql specific files out of the CDK. |
| 3.3.0 | 2023-12-12 | 33018 | Migrate to Per-stream/Global states and away from Legacy states |
| 3.2.1 | 2023-12-11 | 33330 | Parse DatetimeOffset fields with the correct format when used as cursor |
| 3.2.0 | 2023-12-07 | 33225 | CDC : Enable compression of schema history blob in state. |
| 3.1.0 | 2023-11-28 | 32882 | Enforce SSL on Airbyte Cloud. |
| 3.0.2 | 2023-11-27 | 32573 | Format Datetime and Datetime2 datatypes to 6-digit microsecond precision |
| 3.0.1 | 2023-11-22 | 32656 | Adopt java CDK version 0.5.0. |
| 3.0.0 | 2023-11-07 | 31531 | Remapped date, smalldatetime, datetime2, time, and datetimeoffset datatype to their correct Airbyte types |
| 2.0.4 | 2023-11-06 | #32193 | Adopt java CDK version 0.4.1. |
| 2.0.3 | 2023-10-31 | 32024 | Upgrade to Debezium version 2.4.0. |
| 2.0.2 | 2023-10-30 | 31960 | Adopt java CDK version 0.2.0. |
| 2.0.1 | 2023-08-24 | 29821 | Set replication_method display_type to radio, update titles and descriptions, and make CDC the default choice |
| 2.0.0 | 2023-08-22 | 29493 | Set a default cursor for Cdc mode |
| 1.1.1 | 2023-07-24 | 28545 | Support Read Committed snapshot isolation level |
| 1.1.0 | 2023-06-26 | 27737 | License Update: Elv2 |
| 1.0.19 | 2023-06-20 | 27212 | Fix silent exception swallowing in StreamingJdbcDatabase |
| 1.0.18 | 2023-06-14 | 27335 | Remove noisy debug logs |
| 1.0.17 | 2023-05-25 | 26473 | CDC : Limit queue size |
| 1.0.16 | 2023-05-01 | 25740 | Disable index logging |
| 1.0.15 | 2023-04-26 | 25401 | CDC : Upgrade Debezium to version 2.2.0 |
| 1.0.14 | 2023-04-19 | 25345 | Logging : Log database indexes per stream |
| 1.0.13 | 2023-04-19 | 24582 | CDC : refactor for performance improvement |
| 1.0.12 | 2023-04-17 | 25220 | Logging changes : Log additional metadata & clean up noisy logs |
| 1.0.11 | 2023-04-11 | 24656 | CDC minor refactor |
| 1.0.10 | 2023-04-06 | 24820 | Fix data loss bug during an initial failed non-CDC incremental sync |
| 1.0.9 | 2023-04-04 | 24833 | Fix Debezium retry policy configuration |
| 1.0.8 | 2023-03-28 | 24166 | Fix InterruptedException bug during Debezium shutdown |
| 1.0.7 | 2023-03-27 | 24373 | Preparing the connector for CDC checkpointing |
| 1.0.6 | 2023-03-22 | 20760 | Removed redundant date-time datatypes formatting |
| 1.0.5 | 2023-03-21 | 24207 | Fix incorrect schema change warning in CDC mode |
| 1.0.4 | 2023-03-21 | 24275 | Fix error with CDC checkpointing |
| 1.0.3 | 2023-03-15 | 24082 | Fixed NPE during cursor values validation |
| 1.0.2 | 2023-03-14 | 23908 | Log warning on null cursor values |
| 1.0.1 | 2023-03-10 | 23939 | For network isolation, source connector accepts a list of hosts it is allowed to connect |
| 1.0.0 | 2023-03-06 | 23112 | Upgrade Debezium version to 2.1.2 |
| 0.4.29 | 2023-02-24 | 16798 | Add event_serial_no to cdc metadata |
| 0.4.28 | 2023-01-18 | 21348 | Fix error introduced in 18959 in which option initial_waiting_seconds was removed |
| 0.4.27 | 2022-12-14 | 20346 | Consolidate date/time values mapping for JDBC sources |
| 0.4.26 | 2022-12-12 | 18959 | CDC : Don’t timeout if snapshot is not complete. |
| 0.4.25 | 2022-11-04 | 18732 | Upgrade debezium version to 1.9.6 |
| 0.4.24 | 2022-10-25 | 18383 | Better SSH error handling + messages |
| 0.4.23 | 2022-10-21 | 18263 | Fixes bug introduced in 15833 and adds better error messaging for SSH tunnel in Destinations |
| 0.4.22 | 2022-10-19 | 18087 | Better error messaging for configuration errors (SSH configs, choosing an invalid cursor) |
| 0.4.21 | 2022-10-17 | 18041 | Fixes bug introduced 2022-09-12 with SshTunnel, handles iterator exception properly |
| 2022-10-13 | 15535 | Update incremental query to avoid data missing when new data is inserted at the same time as a sync starts under non-CDC incremental mode | |
| 0.4.20 | 2022-09-14 | 15668 | Wrap logs in AirbyteLogMessage |
| 0.4.19 | 2022-09-05 | 16002 | Added ability to specify schemas for discovery during setting connector up |
| 0.4.18 | 2022-09-03 | 14910 | Standardize spec for CDC replication. Replace the replication_method enum with a config object with a method enum field. |
| 0.4.17 | 2022-09-01 | 16261 | Emit state messages more frequently |
| 0.4.16 | 2022-08-18 | 14356 | DB Sources: only show a table can sync incrementally if at least one column can be used as a cursor field |
| 0.4.15 | 2022-08-11 | 15538 | Allow additional properties in db stream state |
| 0.4.14 | 2022-08-10 | 15430 | fixed a bug on handling special character on database name |
| 0.4.13 | 2022-08-04 | 15268 | Added [] enclosing to escape special character in the database name |
| 0.4.12 | 2022-08-02 | 14801 | Fix multiple log bindings |
| 0.4.11 | 2022-07-22 | 14714 | Clarified error message when invalid cursor column selected |
| 0.4.10 | 2022-07-14 | 14574 | Removed additionalProperties:false from JDBC source connectors |
| 0.4.9 | 2022-07-05 | 14379 | Aligned Normal and CDC migration + added some fixes for datatypes processing |
| 0.4.8 | 2022-06-24 | 14121 | Omit using ‘USE’ keyword on Azure SQL with CDC |
| 0.4.5 | 2022-06-23 | 14077 | Use the new state management |
| 0.4.3 | 2022-06-17 | 13887 | Increase version to include changes from 13854 |
| 0.4.2 | 2022-06-06 | 13435 | Adjust JDBC fetch size based on max memory and max row size |
| 0.4.1 | 2022-05-25 | 13419 | Correct enum for Standard method. |
| 0.4.0 | 2022-05-25 | 12759 13168 | For CDC, Add option to ignore existing data and only sync new changes from the database. |
| 0.3.22 | 2022-04-29 | 12480 | Query tables with adaptive fetch size to optimize JDBC memory consumption |
| 0.3.21 | 2022-04-11 | 11729 | Bump mina-sshd from 2.7.0 to 2.8.0 |
| 0.3.19 | 2022-03-31 | 11495 | Adds Support to Chinese MSSQL Server Agent |
| 0.3.18 | 2022-03-29 | 11010 | Adds JDBC Params |
| 0.3.17 | 2022-02-21 | 10242 | Fixed cursor for old connectors that use non-microsecond format. Now connectors work with both formats |
| 0.3.16 | 2022-02-18 | 10242 | Updated timestamp transformation with microseconds |
| 0.3.15 | 2022-02-14 | 10256 | Add -XX:+ExitOnOutOfMemoryError JVM option |
| 0.3.14 | 2022-01-24 | 9554 | Allow handling of java sql date in CDC |
| 0.3.13 | 2022-01-07 | 9094 | Added support for missed data types |
| 0.3.12 | 2021-12-30 | 9206 | Update connector fields title/description |
| 0.3.11 | 2021-12-24 | 8958 | Add support for JdbcType.ARRAY |
| 0.3.10 | 2021-12-01 | 8371 | Fixed incorrect handling “\n” in ssh key |
| 0.3.9 | 2021-11-09 | 7748 | Improve support for binary and varbinary data types |
| 0.3.8 | 2021-10-26 | 7386 | Fixed data type (smalldatetime, smallmoney) conversion from mssql source |
| 0.3.7 | 2021-09-30 | 6585 | Improved SSH Tunnel key generation steps |
| 0.3.6 | 2021-09-17 | 6318 | Added option to connect to DB via SSH |
| 0.3.4 | 2021-08-13 | 4699 | Added json config validator |
| 0.3.3 | 2021-07-05 | 4689 | Add CDC support |
| 0.3.2 | 2021-06-09 | 3973 | Add AIRBYTE_ENTRYPOINT for Kubernetes support |
| 0.3.1 | 2021-06-08 | 3893 | Enable SSL connection |
| 0.3.0 | 2021-04-21 | 2990 | Support namespaces |
| 0.2.3 | 2021-03-28 | 2600 | Add NCHAR and NVCHAR support to DB and cursor type casting |
| 0.2.2 | 2021-03-26 | 2460 | Destination supports destination sync mode |
| 0.2.1 | 2021-03-18 | 2488 | Sources support primary keys |
| 0.2.0 | 2021-03-09 | 2238 | Protocol allows future/unknown properties |
| 0.1.11 | 2021-02-02 | 1887 | Migrate AbstractJdbcSource to use iterators |
| 0.1.10 | 2021-01-25 | 1746 | Fix NPE in State Decorator |
| 0.1.9 | 2021-01-19 | 1724 | Fix JdbcSource handling of tables with same names in different schemas |
| 0.1.9 | 2021-01-14 | 1655 | Fix JdbcSource OOM |
| 0.1.8 | 2021-01-13 | 1588 | Handle invalid numeric values in JDBC source |
| 0.1.6 | 2020-12-09 | 1172 | Support incremental sync |
| 0.1.5 | 2020-11-30 | 1038 | Change JDBC sources to discover more than standard schemas |
| 0.1.4 | 2020-11-30 | 1046 | Add connectors using an index YAML file |