Stop the pod log rotating faster than it can be read

ChoreSecurityService
Shipped
August 23, 2026 at 6:08 AM UTC
Author
Kamo
Commit
a13039f

Roughly 4,600 lines in four minutes, so the retention window was a couple of minutes: chasing a certificate problem on 2026-08-22 the probe output had already rotated away before it could be read, and the evidence had to come from the database instead. Two sources, both noise. **************** makes Hibernate log a twelve-line 'Session Metrics' block at INFO for EVERY session — 172 blocks, ~45% of the volume. The statistics themselves must stay on: Micrometer reads them for hibernate_statements_total, which is how N+1s get found in this service. The logging and the metric are independent, so silence just **************** and keep both. The rest was System.out.println in MemberController — four of them on every member fetch, one of which prints a Hibernate proxy's class name under a comment saying it was added 'to debug instanceof issues'. println has no level, so none of it could be turned off. All 59 are now log.debug (the one that announces a missing avatar URL is log.warn), which keeps them available to anyone who raises the logger and costs nothing otherwise. Also drops Hikari and Spring Cloud Vault from DEBUG to INFO — the pool's housekeeper logged its stats every 30s regardless of activity, and Vault is read once at boot. And application.yml said security: TRACE where prod has always said INFO, so a local run drowned in tracing prod never emits.

All changes

Like what you see shipping?

Every one of these updates lands in your workspace automatically. Start free and watch it grow week after week.

Start Free ForeverView Pricing