Logging that is almost on
Without database-level supplemental logging, the log records the commit but not the row. Capture keeps running and replicates nothing. NeuronCDC's preflight checks the source before the first event.
NeuronCDC reads committed changes straight from the database log and delivers them to your targets in real time — then checks the result, so a missing row becomes an alert, not a surprise at month end.
CAPTURE FROM AND DELIVER TO THE ENGINES YOU ALREADY RUN
Source engines, from Oracle and Tibero to Kafka and flat files
Target types — relational, streaming, NoSQL, search and data lake
Independent checks between a commit and a trusted replica
Agent to deploy — pure Java 17, on Linux, near your data
The agent looks healthy, the lag chart is flat, and rows are still missing. These are the failure modes NeuronCDC was engineered around.
Without database-level supplemental logging, the log records the commit but not the row. Capture keeps running and replicates nothing. NeuronCDC's preflight checks the source before the first event.
A log switch, restart or archive catch-up in the middle of an open transaction is where tools drop work. NeuronCDC resumes before the transaction began, not after it.
A manual truncate or a second writer changes the target behind replication's back. The reconciler and drift validator compare both sides and tell you — or fix it.
Load, capture, apply, verify and fail over with one agent and one repository — no bolt-on tools.
Native readers for Oracle LogMiner and XStream, Tibero redo (including TAC threads), PostgreSQL logical replication, MySQL and MariaDB binlog, SQL Server CDC, Db2 recovery logs and journals, and Altibase ALA.
Copy existing data while capture already runs. Apply for each table is held until its load completes, so the snapshot and the change stream meet exactly once.
A reconciler compares source and target by key and re-applies what is missing. A drift validator runs on a schedule and raises an alarm when counts or values disagree.
Two agents share a lease in the repository. The standby takes over from the last durable checkpoint, and an epoch number fences the old node so a returning process can never write twice.
Changes flow through a durable trail in your own repository database, so capture and apply can stop, restart and fail over independently.
Archive mode, supplemental logging, grants, container names and target tables are checked at start-up. Problems that would otherwise replicate zero rows without an error stop the agent with a clear message.
source.preflight.enabled: trueRename, drop and derive columns, or join lookup tables into a flattened target, using the same code for the snapshot and the stream.
The agent runs inside your network. Built-in SSH tunnels reach databases behind bastions, and nothing leaves for a vendor cloud.
Keyed by primary key, serialised as JSON or Avro with Schema Registry, using transactional producers.
Monitoring endpoint, per-thread lag, sink counters and loss counters — alert on progress, not just liveness.
Every stage writes a durable position, so any component can restart without re-reading from the beginning or skipping ahead.
The capture reader follows redo, WAL or binlog from a saved position and groups records into transactions. Rolled-back work is discarded; only COMMIT moves forward.
Committed events land in the trail tables of your repository — Oracle, PostgreSQL, Tibero and others — with the capture checkpoint in the same transaction.
The applier replays transactions in commit order, maps names and types per target, and marks each event applied. Tables still loading are gated until their snapshot completes.
The reconciler and validator compare source and target on a schedule. Drift is healed or reported, and counters make any gap visible in your monitoring.
See capture and apply lag, trigger a table reload with or without truncate, add a schema or table to a running agent, and read the audit trail — in the browser or with the viriyaneuron_adm CLI.
Load the new platform while the old one serves traffic, keep both in step, and cut over when counts match.
Plan a migrationKeep a warm replica on another engine or site, verified by the reconciler rather than assumed healthy.
Design DRStream production changes into PostgreSQL, Snowflake, or a Hive and Impala data lake as Parquet or Avro, so reports stop competing with transactions.
Offload reportingPublish every row change to Kafka as keyed events and feed microservices, caches and search indexes.
Stream changesNo. Capture reads the database's own change log — redo, WAL, binlog, CDC tables or journals. Source tables are queried only for the initial load and for verification.
ARCHIVELOG mode, database-level supplemental logging, a common user that can run LogMiner in CDB$ROOT, and a PDB user that can read the catalog and the replicated tables. The documentation lists the exact grants.
Yes. Request a reload from Neuron Manager or the CLI. Apply for that table is held while it loads and resumes automatically when the load completes; other tables keep flowing.
On your Linux servers, close to the data, as a single Java 17 process. The trail and checkpoints live in a repository database you choose.
Enable the reconciler and drift validator. They compare source and target by key, repair missing rows, and raise alerts through logs and monitoring metrics.
Bring a source and a target. We will run a proof of concept on your infrastructure and show the verification report.