Loading...

XML

Word

Printable

Type: Bug
Resolution: Done
Priority: Critical
Fix Version/s: 1.0.0.Beta1
Affects Version/s: 0.8.1.Final
Component/s: postgresql-connector
Labels:
- qe-investigate
Environment:
Hide

Oracle Java java 1.8.0_171
Kafka 2.0.0
Debezium Connector Postgres 0.8.1.Final
RDS PostgreSQL 9.6.8

Connector config

name=debezium-db-unstable connector.class=io.debezium.connector.postgresql.PostgresConnector tasks.max=1 database.hostname=postgres-hostname.com database.port=5432 database.user=root database.dbname=unstable database.server.name=unstable table.whitelist=schema.tablename plugin.name=wal2json_rds snapshot.mode=never slot.name=debezium_unstable
Show
Oracle Java java 1.8.0_171 Kafka 2.0.0 Debezium Connector Postgres 0.8.1.Final RDS PostgreSQL 9.6.8 Connector config name=debezium-db-unstable connector.class=io.debezium.connector.postgresql.PostgresConnector tasks.max=1 database.hostname=postgres-hostname.com database.port=5432 database.user=root database.dbname=unstable database.server.name=unstable table.whitelist=schema.tablename plugin.name=wal2json_rds snapshot.mode=never slot.name=debezium_unstable

Git Pull Request:
https://github.com/debezium/debezium/pull/1078, https://github.com/debezium/debezium.github.io/pull/369, https://github.com/debezium/debezium/pull/1082
Steps to Reproduce:

Hide

1. Take a Postgres instance
2. Setup two databases: production, unstable
3. Create a connector capturing events from a table on the unstable database
4. Generate changes a table with the same name in production database
5. See the disk space grow

Show
1. Take a Postgres instance 2. Setup two databases: production, unstable 3. Create a connector capturing events from a table on the unstable database 4. Generate changes a table with the same name in production database 5. See the disk space grow

SFDC Cases Counter:
SFDC Cases Open:
SFDC Cases Links:

After deploying the Debezium connector for Postgres in our unstable environment we faced our Postgres instance quickly running out of space.

We have a Postgres instance where we have two databases: production and unstable.

The connector is configured to capture changes from the unstable database. The changes on unstable are few and the connector works correctly.
However, the traffic on the production database is quite high and we noticed that the disk space increases monotonically until we almost ran out of disk space.

After stopping the connector and deleting the replication slot, the disk space was freed up.
It appears to me that `replication_slots` in Postgres work across databases and the Connector is not listening to the event of other databases.