Hello all,pg_settings ” we have “google_dataplex.max_messages ” with a value like “-1:132 ”
More details can be found here:
opened 05:25PM - 26 Feb 26 UTC
bug
### Description
There is a critical instability in PMM 3.6.0 and 3.5.1 where ad… ding a PostgreSQL service (specifically Cloud SQL) fails to initialize the postgres-exporter. While the "QAN postgresql pgstatements agent" eventually transitions to "Running" after multiple pod restarts, the postgres-exporter remains stuck in a "Starting" state indefinitely.
Attempting to remove the failing service via pmm-admin or the UI causes a total deadlock of the pmm-agent process, rendering the CLI unavailable and requiring a full Kubernetes pod deletion to recover.
<img width="966" height="225" alt="Image" src="https://github.com/user-attachments/assets/a8a23c11-2b5d-4cb7-8b70-779ffd7d75ad" />
### Expected Results
PostgreSQL services should be added successfully and transition to a "Running" state.
### Actual Results
- Permanent "Starting" state for exporters.
- pmm-admin becomes unresponsive during service cleanup.
- Internal logs show connection reset by peer even when credentials and pg_monitor roles are verified.
### Version
PMM server v3.6.0 and 3.5.1
edit: Helm chart version: 1.4.13-1.5.1
### Steps to reproduce
1. Deploy PMM All-in-One (Helm Chart) version 3.5.1 or 3.6.0.
2. Add a PostgreSQL instance via the UI or CLI.
3. Observe the postgres-exporter stuck in Starting.
4. Observe the QAN agent stuck in Waiting (may recover only after 2–3 pod restarts).
5. Attempt pmm-admin remove postgresql <name>.
Result: CLI hangs; pmm-agent stops responding to all commands.
### Relevant logs
```Shell
```
### Code of Conduct
- [x] I agree to follow Percona Community Code of Conduct
Can we expect an updated PostgreSQL exporter version (>= 0.19.0) to be included in an upcoming PMM release? Thanks in advance!