Skip to content
SPECIFICTION
Get listed

TiCDC

docs.pingcap.com

Free plan Open source macOSLinuxAPISelf-hosted
Visit docs.pingcap.com
Spec score 8.1 #6 of 31 in Database Proxy Software

What it is

TiCDC is a free, open-source, self-hosted tool for replicating incremental data changes from TiDB. It captures TiKV change logs, sorts the changes, and exports row-based data to systems including TiDB, MySQL-compatible databases, Kafka, Amazon S3, GCS, Azure Blob Storage, and NFS. Recommended Kafka formats include Canal-JSON, Avro, and Debezium. Filters can select databases, tables, DMLs, and DDLs, while the Open API can check task status, change task configuration, and create or delete tasks. TiCDC supports replication between TiDB clusters for disaster recovery, with second-level RPO and minute-level RTO, as well as bidirectional replication for multi-active deployments. It can be deployed with a new cluster or added to an existing one; production guidance recommends at least two instances. The system has no single point of failure and allows nodes to be added or removed dynamically. Delivery is at least once, so failures can lead to repeated statements or Kafka messages.

Spec sheet

What it does
TiCDC captures TiKV change logs, orders them, and exports row-level incremental changes from TiDB to downstream systems.Read 21 Oct 2026 · docs.pingcap.com
Purpose
TiCDC replicates incremental data from TiDB by capturing TiKV change logs, sorting the changes, and exporting row-based data to downstream systems.Read 2 Oct 2026 · docs.pingcap.com
Intended users
Documented use cases include disaster recovery, real-time data feeds for monitoring, caching, indexing and analysis, and replication between heterogeneous databases.Read 2 Oct 2026 · docs.pingcap.com
Deployment
TiCDC can be deployed with a new TiDB cluster or added to an existing cluster using TiUP.Read 21 Oct 2026 · docs.pingcap.com
Security
The deployment guide documents enabling TLS between TiDB components, including TiCDC.Read 2 Oct 2026 · docs.pingcap.com
Support
PingCAP directs paid-plan users to support tickets and all users to community Slack support.Read 21 Oct 2026 · pingcap.com
Headquarters
Sunnyvale, California, USARead 28 Sep 2026 · docs.pingcap.com
Founded
2015Read 28 Sep 2026 · docs.pingcap.com
API limitation
TiCDC OpenAPI v1 is deprecated, and its task-creation parameters are less complete than the CLI parameters; the documentation recommends OpenAPI v2.Read 21 Oct 2026 · docs.pingcap.com
Availability
TiCDC has no single point of failure and supports dynamically adding and deleting TiCDC nodes.Read 2 Oct 2026 · docs.pingcap.com
Bidirectional replication
TiCDC can replicate changes in both directions between TiDB clusters for multi-active deployments.Read 21 Oct 2026 · docs.pingcap.com
Build requirements
The repository documents compiling TiCDC on Linux or macOS with Go 1.25.12 using make cdc.Read 21 Oct 2026 · github.com
Company founding
LinkedIn states that the TiDB organization was founded in 2015 by three software engineers.Read 21 Oct 2026 · linkedin.com
Company size
LinkedIn lists TiDB powered by PingCAP as having 501–1,000 employees and headquarters in Sunnyvale, California.Read 21 Oct 2026 · linkedin.com
Compatibility requirement
The new architecture requires TiDB 7.5.0 or later.Read 21 Oct 2026 · docs.pingcap.com
Delivery semantics
TiCDC outputs each DDL or DML at least once, so failures can cause repeated statements or Kafka messages.Read 2 Oct 2026 · docs.pingcap.com
Deployment and scale
TiCDC can be deployed with a new TiDB cluster or added to an existing cluster, and the deployment documentation recommends at least two instances in production.Read 2 Oct 2026 · docs.pingcap.com
Disaster recovery
TiCDC supports primary-secondary TiDB replication with second-level recovery-point objectives and minute-level recovery-time objectives.Read 21 Oct 2026 · docs.pingcap.com
Documentation
The TiCDC documentation includes deployment, changefeeds, monitoring, alerting, troubleshooting, performance tuning, integration scenarios, APIs, compatibility, FAQs, and a glossary.Read 21 Oct 2026 · docs.pingcap.com
Download targets
Published binaries cover Linux amd64 and arm64 plus macOS amd64 and arm64.Read 21 Oct 2026 · github.com
Downstream sinks
TiCDC can replicate to TiDB, MySQL-compatible databases, Kafka, Amazon S3, GCS, Azure Blob Storage, and NFS.Read 2 Oct 2026 · docs.pingcap.com
Filtering
Replication rules can filter databases, tables, DML events, and DDL events.Read 21 Oct 2026 · docs.pingcap.com
Filtering and management
TiCDC can filter databases, tables, DMLs, and DDLs, and its Open API can query task status, modify task configuration, and create or delete tasks.Read 2 Oct 2026 · docs.pingcap.com
Funding news
PingCAP announced a $270 million Series D round on November 17, 2020, led jointly by GGV Capital, Access Technology Ventures, Anatole Investment, Jeneration Capital, and 5Y Capital.Read 21 Oct 2026 · prnewswire.co.uk
High availability
TiCDC has no single point of failure and supports dynamically adding or deleting TiCDC nodes.Read 21 Oct 2026 · docs.pingcap.com
Index requirement
TiCDC generally replicates only tables with at least one valid index, though the force-replicate option can replicate tables without one.Read 2 Oct 2026 · docs.pingcap.com
Kafka formats
Recommended Kafka data formats include Canal-JSON, Avro, and Debezium.Read 2 Oct 2026 · docs.pingcap.com
Kafka integration
TiCDC publishes changes to Kafka using Canal-JSON, Avro, Debezium, Open Protocol, or Simple formats.Read 21 Oct 2026 · docs.pingcap.com
Large transactions
TiCDC only partially supports upstream large transactions, and its new architecture has an out-of-memory risk with extremely large transactions.Read 2 Oct 2026 · docs.pingcap.com
Legal company name
Crunchbase lists the legal entity as PingCAP (US) Inc. and identifies Dylan Cui, Edward Huang, and Max Liu as founders.Read 21 Oct 2026 · crunchbase.com
Licence
The public TiCDC repository is licensed under Apache-2.0.Read 21 Oct 2026 · github.com
Limit: cloud operation complexity
The same independent analysis identifies integration cost, service-to-service latency, SLA impact, and operational cost as unresolved cloudification concerns.Read 21 Oct 2026 · amyangfei.me
Limit: delivery semantics
TiCDC emits DDL and DML at least once, so failures can cause repeated SQL statements or duplicate Kafka messages.Read 21 Oct 2026 · docs.pingcap.com
Limit: historical performance
DZone reports that TiCDC previously used substantial CPU for downstream SQL parsing and compiling and suffered from limited plan-cache hits.Read 21 Oct 2026 · dzone.com
Limit: index requirement
Tables normally need a primary key or a qualifying non-null unique index; force-replicate can override this requirement.Read 21 Oct 2026 · docs.pingcap.com
Limit: large transactions
The new architecture does not yet split extremely large transactions into batches, leaving an out-of-memory risk.Read 21 Oct 2026 · docs.pingcap.com
Limit: unsupported scenarios
Documented unsupported cases include RawKV-only clusters, CREATE SEQUENCE handling, and physical TiDB Lightning imports into replicated tables.Read 21 Oct 2026 · docs.pingcap.com
Multi-active
TiCDC supports bidirectional replication between TiDB clusters for multi-active deployments.Read 2 Oct 2026 · docs.pingcap.com
MySQL replication
TiCDC sends low-latency incremental changes to MySQL and other MySQL-compatible databases.Read 21 Oct 2026 · docs.pingcap.com
New architecture performance
The new architecture reports up to 500,000 tables and 190 MiB/s on one node in wide-table scenarios.Read 21 Oct 2026 · docs.pingcap.com
New architecture resource use
PingCAP says typical CPU and memory consumption can fall by up to 50% with the new architecture.Read 21 Oct 2026 · docs.pingcap.com
New architecture scale
The new architecture is documented as scaling beyond 100 nodes, more than 10,000 changefeeds, and millions of tables in one changefeed.Read 21 Oct 2026 · docs.pingcap.com
Open API
The API can query node and cluster status, create, remove, update, pause, and resume replication tasks, rebalance tables, evict owners, and change log levels.Read 21 Oct 2026 · docs.pingcap.com
Praise: component design
An independent TiCDC engineering analysis describes puller, sorter, mounter, and sink components as independently focused with high cohesion and low coupling.Read 21 Oct 2026 · amyangfei.me
Security: client authentication
Since version 8.1.0, TiCDC supports mTLS authentication and TiDB username/password authentication, independently or together.Read 21 Oct 2026 · docs.pingcap.com
Security: TLS warning
The documentation warns that username and password are plaintext when TLS is disabled, creating a credential-leak risk.Read 21 Oct 2026 · docs.pingcap.com
Storage integrations
Supported storage destinations include Amazon S3, Google Cloud Storage, Azure Blob Storage, and NFS.Read 21 Oct 2026 · docs.pingcap.com
System requirements
Production guidance recommends at least 16 CPU cores, 64 GB memory, a 500 GB SSD, a 10-Gigabit network card, and two TiCDC instances.Read 21 Oct 2026 · docs.pingcap.com
Unsupported cases
The docs list RawKV-only clusters, CREATE SEQUENCE and SEQUENCE functionality, and TiDB Lightning physical import on tables being replicated as unsupported scenarios.Read 2 Oct 2026 · docs.pingcap.com

Plans and prices

TiCDC (open-source) Free No paid plans or usage caps stated on the TiCDC pages opened Read 2 Oct 2026 · docs.pingcap.com

Every TiCDC plan and price →

Strong points and limits

Strong points

  • Replicates to databases, Kafka, and cloud storage.
  • Filters databases, tables, DMLs, and DDLs.
  • Supports bidirectional TiDB cluster replication.
  • Open-source and free.

Limits

  • Delivery is at least once, so duplicates can occur after failures.
  • Generally requires a valid table index.
  • Extremely large transactions can pose an out-of-memory risk.
  • Production guidance recommends at least two instances.

Who it suits

TiCDC suits teams that operate TiDB and need change-data feeds, cluster replication, or downstream data for monitoring, caching, indexing, or analysis. It is self-hosted and can support replication to heterogeneous databases.

Verdict

TiCDC covers several replication destinations and provides task controls through its Open API. Review its delivery behavior, index requirements, and documented unsupported cases before relying on it.

Where Specifiction places it

Alternatives to TiCDCAll alternatives →