Ory is the largest open source community in the world for cloud software application security. We maintain advanced open source security software solving authentication, authorization, access control, application network security, and delegation. Ory implements a variety of industry and best-practice standards including OAuth 2.0 / OAuth 2.1, OpenID Connect, Zero Trust Networking, Google Zanzibar Policy Framework, FIDO2 U2F, WebAuthn, TOTP, and more.

Ory Community

Within Kratos, does the postgres client used have any mechanism to retry/detected a failed connection? I'm not sure if connection pooling is used at all? We're seeing intermitent connection failures, only with Kratos and I'm trying to understand why.

Kratos on the rare occasion returns:

```unable to fetch records: write failed: write tcp 127.0.0.1:58156-127.0.0.1:5432: write: broken pipe```
The postgres instance runs within GCP CloudSQL, and shows no errors in its logs. The same instance is also shared with other services which report no connection issues at the same time either, so I'm confident to rule out a Postgres issue (performance, connection limits are all healthy too)

We do run the CloudSQL Auth Proxy as a sidecar container for all,  which reports the below error

```connection aborted - error reading from client: read tcp 127.0.0.1:5432-&gt;127.0.0.1:41954: read: connection reset by peer```

Reviewing the logs for all services that we run within Kubernetes, (which share postgres, and run identical cloudsql proxies), we only observe these errors within Kratos.

Any pointers?