Logical replication creates infinite loop of inserts

[marceloneppel]

Steps to reproduce

1. juju deploy postgresql-k8s --channel 16/edge --trust postgresql1
2. juju deploy postgresql-k8s --channel 16/edge --trust postgresql2
3. juju deploy data-integrator di1 --config database-name=testdb
4. juju deploy data-integrator di2 --config database-name=testdb
5. juju integrate postgresql1 di1
6. juju integrate postgresql2 di2
7. juju integrate postgresql1:logical-replication-offer postgresql2:logical-replication
8. juju integrate postgresql1:logical-replication postgresql2:logical-replication-offer
9. juju run di1/leader get-credentials
10. psql PSQL URI FROM ABOVE
11. create table asd (message int);\q
12. juju run di2/leader get-credentials
13. psql PSQL URI FROM ABOVE
14. create table asd (message int);\q
15. juju config postgresql1 logical_replication_subscription_request='{"testdb": ["public.asd"]}'
16. juju config postgresql2 logical_replication_subscription_request='{"testdb": ["public.asd"]}'

Expected behaviour

A blocked status message telling that the same table cannot be defined on both sides of the relation for the logical replication.

Actual behaviour

An infinite loop of inserts happens due to the cyclic replication that was configured by configuring the same table on both sides of the relation.

Versions

Operating system: Ubuntu 24.04

Juju CLI: 3.6.8

Juju agent: 3.6.8

Charm revision: 638

microk8s: 1.32-strict/stable

Log output

Juju debug log: no relevant logs.

Additional context

The table needs to be empty in both clusters; otherwise, the charm will properly set a blocked status indicating the table is not empty in one of the clusters.

This is the same situation as reported at canonical/postgresql-operator#1085.

[syncronize-issues-to-jira]

Thank you for reporting your feedback to us!

The internal ticket has been created: https://warthogs.atlassian.net/browse/DPE-7883.

This message was autogenerated