Different tasks may require extra configs, refer to task templates and tutorial

For configuration changes between releases, see Config changelog.

[extractor]

Config	Description	Example	Default
db_type	source database type	mysql	required
extract_type	extraction type; available values depend on db_type	snapshot	required
url	database URL; credentials may be included in the URL or configured separately	mysql://127.0.0.1:3307	empty
username	database connection username	root	empty
password	database connection password	password	empty
ssl_mode	MySQL/PostgreSQL/MSSQL/Redis/MongoDB TLS mode: disable, require, verify_ca, verify_full.	verify_ca	not set
ssl_ca_path	CA certificate path used by MySQL/PostgreSQL/MSSQL/Redis/MongoDB TLS verification	/etc/ssl/certs/ca.pem	empty
ssl_client_cert_path	Client certificate PEM path (MongoDB: combined certificate/private key PEM); unsupported by MSSQL	/etc/ssl/client.crt	empty
ssl_client_key_path	Client private key PEM path; empty uses ssl_client_cert_path as a combined PEM; MongoDB requires the combined file	/etc/ssl/client.key	empty
max_connections	maximum source connection pool size	10	10
batch_size	number of rows extracted per batch; if using chunk splitting, this is also the target chunk size for the source	10000	[pipeline].buffer_size / effective snapshot parallel_size. If set to 0, uses [pipeline].buffer_size directly
max_rps	optional source-side rate limit in records per second; 0 disables the limit	1000	0
max_mbps	optional source-side rate limit in MiB per second; 0 disables the limit	100	0
app_name	connection application name, currently used by MongoDB	APE_DTS	APE_DTS
parallel_type	snapshot extraction parallel strategy	table	table
parallel_size	source snapshot worker limit	4	1; legacy fallback: [runtime].tb_parallel_size
partition_cols	partition column for data splitting during MySQL/PostgreSQL snapshot migration; only one column per table is supported	json:[{"db":"db_1","tb":"tb_1","partition_col":"id"},{"db":"db_2","tb":"tb_2","partition_col":"id"}]	empty
is_direct_connection	MongoDB driver directConnection option	true	not set (driver default)
is_cluster	Redis Cluster mode for snapshot/CDC/snapshot-and-CDC	true	not set or empty (detect from the connected Redis node)

URL escaping

• If the username/password contains special characters, the corresponding parts need to be percent-encoded, for example:

create user user1@'%' identified by 'abc%$#?@';
The url should be:
url=mysql://user1:abc%25%24%23%3F%40@127.0.0.1:3307?ssl-mode=disabled

Credentials configured through username and password are percent-encoded and merged into the URL by DTS. If ssl_mode is set, ssl_ca_path is optional unless the selected verification mode and server setup require a CA certificate.

TLS Options

ssl_mode controls the client's TLS requirements and verification of the server:

• disable: plaintext without TLS.
• require: encryption without verifying the server certificate or hostname.
• verify_ca: encryption and server certificate-chain verification against a trusted CA.
• verify_full: the checks in verify_ca, plus verification of the URL hostname/IP against the certificate SAN.

Client-certificate authentication is a separate server policy. An optional client identity can be configured with any encrypted mode; verify_full does not require one by itself.

ssl_client_cert_path and ssl_client_key_path configure a PEM client identity for MySQL, PostgreSQL and Redis, including metadata and checkpoint connections. MySQL CDC is an exception: its current binlog driver supports only disable/require and no client certificates; unsupported inputs are rejected without downgrading. An empty key path uses the same combined PEM as the certificate path. MongoDB requires a combined PEM in ssl_client_cert_path and an empty ssl_client_key_path. MSSQL does not support client certificates and rejects those inputs.

Driver limits for verify_ca:

• Redis and PostgreSQL CDC skip hostname checks while retaining certificate-chain verification.
• The current SQLx rustls driver maps verify_ca but does not recognize rustls's newer NotValidForNameContext error. MySQL/PostgreSQL query connections therefore still reject hostname mismatches in this mode.
• MongoDB's rustls backend and Tiberius (MSSQL) have no independent hostname opt-out. Their verify_ca currently performs the same server checks as verify_full.

MSSQL supports all four modes. Its two verification modes require ssl_ca_path; explicit task SSL settings override URL, ADO.NET and JDBC encryption/trust settings.

Task Support

This matrix describes the five engines covered by these TLS settings on this branch. The driver-specific limits for verify_ca are described above.

Engine	ssl_mode	struct	snapshot	CDC	checker
MySQL	disable, require	Yes	Yes	Yes	Yes
MySQL	verify_ca, verify_full	Yes	Yes	No (binlog driver limitation)	Yes
PostgreSQL	disable, require, verify_ca, verify_full	Yes	Yes	Yes	Yes
MSSQL	disable, require, verify_ca, verify_full	No	Yes	No	No
Redis	disable, require, verify_ca, verify_full	No	Yes	Yes (including snapshot-and-CDC)	No
MongoDB	disable, require, verify_ca, verify_full	Yes	Yes	Yes	Yes

TLS tests and fixtures share dt-tests/tests/tls/, with sibling engine/topology/version directories (for example, mongo_shard and redis_cluster_6_2). Each directory groups fixtures by task, while the shared test harness varies SSL parameters without copying prepare data for each SSL mode. The runner uses three independent suites: tls (MySQL/PostgreSQL/MSSQL), mongo_to_mongo_tls, and redis_to_redis_tls. The tls suite covers all supported relational cells above. Redis TLS tests exercise snapshot-and-CDC, including cluster encryption; MongoDB TLS tests exercise snapshot/struct/CDC with require, including sharding, plus a verify_full snapshot. MongoDB checker and verified cluster workflows reuse the same client mapping but do not have a dedicated TLS E2E test in this change.

TLS task accounts and servers require trusted client certificates where supported. MySQL CDC uses a separate source/target container pair that requires encryption without client certificates; MSSQL does not support TLS client certificates.

Redis TLS

• Redis URLs support redis:// and rediss://. Without ssl_mode, redis:// is plaintext and rediss:// uses TLS without server certificate verification.
• Redis supports disable, require, verify_ca, and verify_full. Both verification modes require ssl_ca_path: verify_ca checks the CA chain, and verify_full additionally checks the URL hostname/IP against the certificate SAN.
• An explicit ssl_mode overrides the URL scheme and fragment. A DNS URL host is sent as SNI for both verification modes.
• These rules apply to ordinary Redis command connections and PSYNC replication streams, and are preserved for discovered Redis Cluster node URLs.
• With Redis Cluster and either verification mode, every node must present a certificate signed by the configured CA. With verify_full, each discovered node hostname or IP must also match its certificate SAN.

MongoDB TLS

• MongoDB uses the driver's rustls backend. Without ssl_mode, TLS options come from the MongoDB URI, including the driver's mongodb+srv:// defaults.
• disable turns TLS off. require encrypts the connection without verifying the server certificate or hostname; no CA file is needed.
• Both verify_ca and verify_full require ssl_ca_path and verify the certificate chain and hostname with rustls. These two modes have the same behavior for MongoDB: the URI hostname or IP must match the server certificate SAN.
• An explicit ssl_mode replaces the URI TLS options. These settings apply to extractor, sinker, and database checkpoint connections, including replica sets and sharded clusters.

extractor.parallel_type

• table: allocate snapshot concurrency across tables. With parallel_size=4, up to 4 tables can be extracted at the same time.
• chunk: allocate snapshot concurrency within a single table by chunk splitting. With parallel_size=4, one table can run up to 4 chunk workers in parallel.
• When parallel_type=chunk, [extractor].batch_size is also the target chunk size. Chunk boundaries are data-dependent, so the actual row count may differ, but the extractor tries to make each chunk close to batch_size.
• parallel_size is the effective concurrency limit in both modes.
• MySQL and PostgreSQL snapshot extractors support both table and chunk.
• MongoDB snapshot extractors currently support only table; chunk is not supported.
• Deprecated compatibility: [runtime] tb_parallel_size is kept only as a legacy fallback when [extractor] parallel_size is not set.

Redis source cluster mode

• [extractor].url can point to any reachable node in the source cluster. DTS discovers all source master nodes through CLUSTER NODES and starts one PSYNC extractor for each master.
• [extractor].is_cluster is optional. When omitted, DTS connects to the Redis node specified by [extractor].url and detects whether Redis Cluster mode should be used from the node's actual cluster state.
• Set [extractor].is_cluster=true to force Redis Cluster mode. DTS discovers and syncs the whole source cluster.
• Set [extractor].is_cluster=false to force single-node Redis mode. DTS runs PSYNC only against the node specified by [extractor].url. This can be used when the source is a Redis Cluster but only one cluster node should be synced.

Mongo source connection mode

• [extractor].is_direct_connection maps to the MongoDB driver directConnection option.
• Omit it to let the driver infer the topology from the URL. This is the recommended default for replica sets and sharded clusters.
• Set it only when you intentionally want to connect directly to a specific MongoDB node. Do not set it to true when connecting through mongos for sharded-cluster CDC or snapshot tasks.

[sinker]

Config	Description	Example	Default
db_type	target database type	mysql	required except for sink_type=dummy
sink_type	target operation; supported values depend on db_type	write	write when [sinker] exists; dummy when omitted for standalone checker
url	database URL; credentials may be included in the URL or configured separately	mysql://127.0.0.1:3307	empty
username	database connection username	root	empty
password	database connection password	password	empty
ssl_mode	MySQL/PostgreSQL/MSSQL/Redis/MongoDB TLS mode: disable, require, verify_ca, verify_full.	verify_ca	not set
ssl_ca_path	CA certificate path used by MySQL/PostgreSQL/MSSQL/Redis/MongoDB TLS verification	/etc/ssl/certs/ca.pem	empty
ssl_client_cert_path	Client certificate PEM path (MongoDB: combined certificate/private key PEM); unsupported by MSSQL	/etc/ssl/client.crt	empty
ssl_client_key_path	Client private key PEM path; empty uses ssl_client_cert_path as a combined PEM; MongoDB requires the combined file	/etc/ssl/client.key	empty
max_connections	maximum target connection pool size	10	10
batch_size	records written per batch; must be greater than 0	200	200
max_rps	optional target-side rate limit in records per second; 0 disables the limit	1000	0
max_mbps	optional target-side rate limit in MiB per second; 0 disables the limit	100	0
replace	replace an existing row on insert conflict, for MySQL/PostgreSQL snapshot and CDC tasks	false	true
disable_foreign_key_checks	disable foreign-key checks while writing MySQL/PostgreSQL	true	true
transaction_isolation	MySQL/TiDB target transaction isolation: default, read_uncommitted, read_committed, repeatable_read, or serializable	read_committed	default
conflict_policy	structure migration conflict policy: interrupt or ignore	interrupt	interrupt
app_name	connection application name, currently used by MongoDB	APE_DTS	APE_DTS
is_direct_connection	MongoDB driver directConnection option	true	not set (driver default)
is_cluster	Redis Cluster mode	true	not set or empty (detect from the connected Redis node)
mongo_require_shard_key_filter	fail fast when a MongoDB update/delete/upsert filter cannot contain the complete target shard key	true	true

Redis target cluster mode

• [sinker].url can point to any reachable node in the target cluster. DTS discovers all target master nodes through CLUSTER NODES and routes Redis commands to the owning node by key slot.
• In Redis target cluster mode, DTS creates sinkers according to the target master nodes, instead of limiting the sinker count by [parallelizer].parallel_size.
• [sinker].is_cluster is optional. When omitted, DTS connects to the Redis node specified by [sinker].url and detects whether Redis Cluster mode should be used from the node's actual cluster state.
• Set [sinker].is_cluster=true to force Redis Cluster mode when writing to the target cluster.
• Set [sinker].is_cluster=false to force single-node Redis mode and write only to the node specified by [sinker].url.

Mongo target connection and shard-key mode

• [sinker].is_direct_connection maps to the MongoDB driver directConnection option. Omit it to let the driver infer the topology from the URL. For sharded targets, connect through mongos and do not set it to true.
• [sinker].mongo_require_shard_key_filter=true is the default. When the target collection is sharded, DTS checks whether update/delete/upsert filters contain the full target shard key and fails fast if required shard key fields are missing.
• Keep mongo_require_shard_key_filter=true for normal migrations. Set it to false only when you explicitly accept MongoDB server-side routing behavior, such as a controlled best-effort migration on a compatible MongoDB version.

[checker]

Common row/structure comparison settings. The section is used in these modes:

• Standalone snapshot/struct/check-log: set [sinker].sink_type=check. The target connection, authentication, TLS, connection limits, and database-specific options are all loaded from [sinker].
• Inline snapshot: use extract_type=snapshot, [sinker].sink_type=write, and add a [checker] section. Checking runs synchronously after each successful sink operation.
• Inline CDC: use extract_type=cdc, [sinker].sink_type=write, and enable [checker_cdc].is_enabled=true. Checking runs asynchronously after sink through the CDC checker queue.

Config	Description	Example	Default
batch_size	maximum rows processed by one checker query	200	200
sample_percent	percentage sampled for snapshot/CDC checks	25	empty (check every row/change)
recheck_count	number of retries for a temporary inconsistency	4	0
recheck_interval_secs	interval between retries, in seconds	5	0
recheck_queue_size	maximum pending rows in the retry buffer	10000	10000
recheck_queue_memory_mb	retry-buffer memory limit in MiB	256	256

Notes:

• Checker tasks support only [pipeline].pipeline_type=basic.
• sample_percent accepts 1..=100 and applies only to snapshot checks and inline CDC checks. Standalone snapshot applies sampling during extraction. Inline snapshot/CDC writes every row/change and applies deterministic key-hash sampling before target fetch.
• Standalone snapshot check supports MySQL, PostgreSQL, and MongoDB targets. Standalone struct check supports MySQL and PostgreSQL.
• Inline snapshot check supports MySQL, PostgreSQL, and MongoDB write targets.
• recheck_count and recheck_interval_secs are not used by inline CDC reconciliation.
• When either retry-buffer limit is reached, the checker does not drop the result; newly found inconsistencies skip retry and are finalized immediately.

Standalone target example

[extractor]
db_type=mysql
extract_type=snapshot
url=mysql://source-host:3306

[sinker]
db_type=mysql
sink_type=check
url=mysql://target-host:3306
username=root
password=target-password
max_connections=8

[checker]
batch_size=200
sample_percent=25
recheck_count=4
recheck_interval_secs=5
recheck_queue_size=10000
recheck_queue_memory_mb=256

[checker_output]

Check-result output configuration. If this section is omitted, results are written as local logs under runtime.log_dir/check.

Config	Description	Example	Default
output_type	result destination: logs or s3	logs	logs
output_full_row	include complete source/target rows in difference logs	false	false
output_revise_sql	generate repair statements in sql.log	true	false
revise_match_full_row	use the complete row in generated repair predicates	false	false
check_log_dir	local check-log directory	/tmp/check	empty (use runtime.log_dir/check)
check_log_file_size	per-file size limit for diff.log, miss.log, and sql.log	100mb	100mb
check_log_max_rows	maximum rows in diff.log/miss.log	1000	1000
s3_bucket	S3 bucket; required for output_type=s3	my-bucket	-
s3_access_key_id	S3 access key	AKIA...	empty
s3_secret_access_key	S3 secret key	****	empty
s3_region	S3 region	us-east-1	empty
s3_endpoint	custom S3 endpoint	https://...	empty
s3_root_dir	local/mounted root used by the S3 helper	/tmp/check	empty
s3_root_url	root URL used by the S3 helper	s3://bucket	empty
s3_key_prefix	key prefix for uploaded check logs	task1/check	empty

output_type=s3 is supported for standalone snapshot check and inline CDC check. S3 output still uses the configured local rolling-log directory and limits before upload. Structure check, check-log review, and inline snapshot check support output_type=logs only.

[checker_cdc]

CDC-only asynchronous checker settings.

Config	Description	Example	Default
is_enabled	enable inline CDC check	true	false
queue_size	pending CDC checker batches	200	200
check_log_interval_secs	periodic CDC check-result output interval in seconds	30	30

Inline CDC check additionally requires:

• [extractor].extract_type=cdc
• [sinker].sink_type=write with a MySQL/PostgreSQL target
• [parallelizer].parallel_type=rdb_merge
• [resumer].resume_type=from_target or from_db

The CDC checker queue is deliberately decoupled from the migration pipeline. When full, it evicts the oldest pending batch instead of blocking writes. Checker processing/output failures are logged without failing the main CDC write path. Snapshot and structure checks do not use this queue.

[filter]

Config	Description	Example	Default
do_dbs	databases to be synced, takes union with do_tbs	db_1,db_2*,db*&#	-
ignore_dbs	databases to be filtered, takes union with ignore_tbs	db_1,db_2*,db*&#	-
do_tbs	tables to be synced, takes union with do_dbs	db_1.tb_1,db_2*.tb_2*,db*&#.tb*&#	-
ignore_tbs	tables to be filtered, takes union with ignore_dbs	db_1.tb_1,db_2*.tb_2*,db*&#.tb*&#	-
ignore_cols	table columns to be filtered	json:[{"db":"db_1","tb":"tb_1","ignore_cols":["f_2","f_3"]},{"db":"db_2","tb":"tb_2","ignore_cols":["f_3"]}]	-
do_events	events to be synced	insert,update,delete	*
do_ddls	ddls to be synced, for mysql cdc tasks	create_database,drop_database,alter_database,create_table,drop_table,truncate_table,rename_table,alter_table,create_index,drop_index	-
do_dcls	DCL statements to be synced, for supported structure tasks	create_user,grant	-
do_structures	structures to be migrated in structure migration tasks	mysql/pg: database,table,constraint,sequence,comment,index; mongo: collection,shardkey	*
ignore_cmds	commands to be filtered, for redis cdc tasks	flushall,flushdb	-
where_conditions	where conditions for the source SELECT SQL during snapshot migration	json:[{"db":"db_1","tb":"tb_1","condition":"f_0 > 1"},{"db":"db_2","tb":"tb_2","condition":"f_0 > 1 AND f_1 < 9"}]	-

Values

• All configurations support multiple items, which are separated by ",". Example: do_dbs=db_1,db_2.
• Set to * to match all. Example: do_dbs=*.
• Keep empty to match nothing. Example: ignore_dbs=.
• ignore_cols and where_conditions are in JSON format and must start with json:.
• do_events takes one or more values from insert, update, and delete.
• do_dcls takes one or more values from create_user, alter_user, create_role, drop_user, drop_role, grant, revoke, and set_role.
• do_structures takes structure object types. For MySQL/PostgreSQL, common values include database, table, constraint, sequence, comment, and index. For MongoDB, supported values are collection, shardkey. MongoDB does not use a separate database structure type; databases are created implicitly by creating collections. shardkey copies source sharding definitions for sharded collections and runs only when the target is connected through mongos.

Priority

• ignore_tbs + ignore_dbs > do_tbs + do_dbs.
• If a table matches both ignore configs and do configs, the table will be filtered.
• If both do_tbs and do_dbs are configured, the filter is the union of both. If both ignore_tbs and ignore_dbs are configured, the filter is the union of both.

Wildcard

Wildcard	Description
*	Matches multiple characters
?	Matches 0 or 1 characters

Used in: do_dbs, ignore_dbs, do_tbs, and ignore_tbs.

Escapes

Database	Before	After
mysql	db*&#	`db*&#`
mysql	db*&#.tb*$#	`db*&#`.`tb*$#`
pg	db*&#	"db*&#"
pg	db*&#.tb*$#	"db*&#"."tb*$#"

Names should be enclosed in escape characters if there are special characters.

Used in: do_dbs, ignore_dbs, do_tbs and ignore_tbs.

[router]

Config	Description	Example	Default
db_map	database mapping	db_1:dst_db_1,db_2:dst_db_2	-
tb_map	table mapping	db_1.tb_1:dst_db_1.dst_tb_1,db_1.tb_2:dst_db_1.dst_tb_2	-
col_map	column mapping	json:[{"db":"db_1","tb":"tb_1","col_map":{"f_0":"dst_f_0","f_1":"dst_f_1"}}]	-
topic_map	table -> kafka topic mapping, for mysql/pg -> kafka tasks. required	*.*:default_topic,test_db_2.*:topic2,test_db_2.tb_1:topic3	-

Values

• A mapping rule consists of the source and target, which are separated by ":".
• All configurations support multiple items, which are separated by ",". Example: db_map=db_1:dst_db_1,db_2:dst_db_2.
• col_map value is in JSON format and must start with json:.
• If not set, data will be routed to the same databases/tables/columns with the source database.

Priority

• tb_map > db_map.
• col_map only works for column mapping. If a table needs database + table + column mapping, tb_map/db_map must be set.
• topic_map: test_db_2.tb_1:topic3 > test_db_2.*:topic2 > *.*:default_topic.

Wildcard

Not supported.

Escapes

Same with [filter].

[pipeline]

Config	Description	Example	Default
buffer_size	max cached records in memory	16000	16000
buffer_memory_mb	[optional] memory limit for buffer, if reached, new records will be blocked even if buffer_size is not reached, 0 means not set	200	0
checkpoint_interval_secs	interval to flush logs/statistics/position	10	10
batch_sink_interval_secs	maximum interval before flushing a non-empty sink batch	1	0
counter_time_window_secs	time window for monitor counters	10	same with [pipeline] checkpoint_interval_secs
counter_max_sub_count	maximum number of sub-counters	1000	1000
pipeline_type	pipeline implementation; only basic is supported	basic	basic

[parallelizer]

Config	Description	Example	Default
parallel_type	parallel type	snapshot	serial
parallel_size	threads for parallel syncing	8	1
rebalance_strategy	snapshot chunk rebalance strategy used during sink writes	none	none
rebalance_cost	cost metric used to measure partition size	rows	rows
rebalance_max_partitions_per_sinker	max split partitions per effective sinker	2	2
rebalance_min_partition_rows	minimum rows kept in each split snapshot insert partition	200	[sinker].batch_size
rebalance_split_skew_ratio	skew threshold used by the auto_split strategy	1.0	1.0

parallel_type

Type	Strategy	Usage	Advantages	Disadvantages
snapshot	Records in cache are divided into [parallel_size] partitions, and each partition will be synced in batches in a separate thread.	snapshot tasks for mysql/pg/mongo	fast
serial	Single thread, one by one.	all		slow
rdb_merge	Merge row changes in cache into write-friendly insert + delete batches, then divide them into [parallel_size] partitions for parallel syncing. It is used by MySQL/PG CDC, check, review, and revise flows.	mysql/pg CDC, check, review, revise	fast	eventual consistency
mongo	Mongo version of merge parallelization, also used by standalone MongoDB check and review flows.	mongo CDC, check, review
redis	Single thread, batch/serial writing(determined by [sinker] batch_size)	snapshot/CDC tasks for redis

snapshot chunk rebalance

When [parallelizer].parallel_type=snapshot, snapshot parallelizer uses chunk partitioner to rebalance the downstream write queue. It is mainly for snapshot write tasks and reduces sink-side long tails. It does not change source-side extractor concurrency and does not rewrite checkpoint chunk ids.

Default behavior:

[parallelizer]
parallel_type=snapshot
parallel_size=8
rebalance_strategy=none
rebalance_cost=rows

The default rebalance_strategy=none keeps logical chunk order after grouping and does not add sink-side sorting or splitting. If sink-side long tails are obvious, use rebalance_strategy=auto_split. Use table_min_rows or table_even for rows-only table-level partitioning. Use the default rebalance_cost=rows when row width is similar. If rows contain large JSON, LOB, or wide strings, use rebalance_cost=bytes. If the target has high request overhead, or you do not want to split logical chunks, use rebalance_strategy=chunk_largest_first.

For scenario-based tuning, see Snapshot Chunk Partitioner Rebalance.

[runtime]

Config	Description	Example	Default
log_level	level	info/warn/error/debug/trace	info
log4rs_file	log4rs config file	./log4rs.yaml	./log4rs.yaml
log_dir	output dir	./logs	./logs
check_result_stdout_only	output only check result logs to stdout	true/false	false

Note that the log files contain progress information for the task, which can be used for task resuming at breakpoint. Therefore, if you have multiple tasks, please set up separate log directories for each task.

[global]

Config	Description	Example	Default
task_id	Unique task identifier	cdc_task_1

In some scenarios, task_id is used to distinguish task uniqueness, such as when using resumer from database. By default, it will be automatically generated based on key configuration information.

[resumer]

Config	Description	Example	Default
resume_type	dummy, from_log, from_target, or from_db	from_target	dummy
log_dir	log directory used by from_log	./logs	[runtime].log_dir
config_file	optional resume config file used by from_log	./resume.config	empty
url	database URL used by from_db	mysql://127.0.0.1:3306	required for from_db
db_type	database type used by from_db	mysql	required for from_db
username	database username used by from_db	root	empty
password	database password used by from_db	password	empty
ssl_mode	MySQL/PostgreSQL/Redis/MongoDB TLS mode used by from_db: disable, require, verify_ca, verify_full.	verify_ca	not set
ssl_ca_path	CA certificate path used by MySQL/PostgreSQL/Redis/MongoDB from_db TLS verification	/etc/ssl/certs/ca.pem	empty
ssl_client_cert_path	Client certificate PEM path (MongoDB: combined certificate/private key PEM); unsupported by MSSQL	/etc/ssl/client.crt	empty
ssl_client_key_path	Client private key PEM path; empty uses ssl_client_cert_path as a combined PEM; MongoDB requires the combined file	/etc/ssl/client.key	empty
is_direct_connection	MongoDB driver directConnection option used by from_db	true	not set
table_full_name	target table used to store resume state for from_db or from_target	apecloud_metadata.apedts_task_position	empty
max_connections	maximum resumer connection pool size	5	5

For details, please refer to the resumer documentation: resuming at breakpoint.

resume_type=from_target reuses the parsed sinker target. For a standalone checker with a dummy or omitted sinker, it reuses the checker target. The legacy keys resume_from_log, resume_log_dir, and resume_config_file are rejected; migrate them to resume_type=from_log, log_dir, and config_file.

[tracing]

Config	Description	Example	Default
task_summary_mode	trace aggregation mode: task or marker	marker	marker
output_format	trace output format: plain or json	json	plain

The runtime trace summary is dumped periodically (every pipeline.checkpoint_interval_secs) to the runtime trace log, including a final dump on task shutdown, so long-running CDC tasks get continuous diagnostics. When both metrics and tracing features are enabled, per-marker task counters and globally aggregated per-wait-point counters (runtime_trace_*) are also exposed on the Prometheus /metrics endpoint. In task summary mode, completed task details are emitted once on the next dump and then released; marker summaries and Prometheus counters remain cumulative.

[metacenter]

This optional section is used by the MySQL dbengine metadata-center mode.

Config	Description	Example	Default
type	metadata-center type: basic or dbengine	dbengine	basic
url	metadata database URL; required for MySQL dbengine mode	mysql://127.0.0.1:3306	required
username	metadata database username	root	empty
password	metadata database password	password	empty
ssl_mode	MySQL TLS mode: disable, require, verify_ca, verify_full.	verify_full	not set
ssl_ca_path	CA certificate path	/etc/ssl/certs/ca.pem	empty
ssl_client_cert_path	Client certificate PEM path (MongoDB: combined certificate/private key PEM); unsupported by MSSQL	/etc/ssl/client.crt	empty
ssl_client_key_path	Client private key PEM path; empty uses ssl_client_cert_path as a combined PEM; MongoDB requires the combined file	/etc/ssl/client.key	empty
ddl_conflict_policy	DDL conflict policy: interrupt or ignore	interrupt	interrupt

The metadata-center URL must differ from both the extractor URL and the effective destination URL.

[data_marker]

If this section is present, the required topology marker configuration is loaded.

Config	Description	Default
topo_name	topology name	required
topo_nodes	topology node list	empty
src_node	source node	required
dst_node	destination node	required
do_nodes	included nodes	required
ignore_nodes	excluded nodes	empty
marker	marker value	required

[processor]

Config	Description	Default
lua_code_file	Lua processor source file loaded by DTS	empty

[metrics]

This section is available only when DTS is built with the metrics feature.

Config	Description	Example	Default
http_host	metrics HTTP bind address	0.0.0.0	0.0.0.0
http_port	metrics HTTP port	9090	9090
workers	metrics HTTP worker count	2	2
labels	comma-separated key=value metric labels	env=prod,az=a	empty