TL;DR'A logging bug may write TBs to local SSDs' — a June 2026 HN risk report at 510 points. Mind the may in the title: it's a warning sign, not a confirmed verdict. This piece turns it into a run-it-yourself SOP plus one two-minute monthly habit.
Sources SOURCES
Distilled from public community discussions. This page only breaks them down and verifies — it never reposts the original; opinions belong to their authors.
Facts FACTS
Check these claims against the listed sources. For allowance mechanics, first-party account displays outrank community summaries.
- Around 2026-06-22, the GitHub issue 'Codex logging bug may write TBs to local SSDs' appeared; the HN discussion pulled 510 points and 271 comments — a disk silently filling up is a failure mode everyone can picture a little too well.
- The issue title says may: this is a risk report, not a confirmed widespread fault. Whether your machine is affected can only be answered by actually running du and df on it.
- The event is logged in this site's known-events archive alongside sub-agent prompt encryption and Codex bypassing sudo — the mid-2026 batch of community scrutiny on agent behavior boundaries.
Concepts CONCEPTS
Not everything said in a community thread is a fact — each concept carries its nature label. Read the label first.
Diagnostic and session logs keep appending as you use the tool; without rotation or cleanup, usage only trends up. Not a Codex-specific problem, but agent tools generate far more log volume per task than traditional CLIs, so bloat arrives faster.
The order of magnitude the issue reports: cumulative logs possibly reaching TBs. Whether that scale is typical, and by what mechanism, is up to the issue thread and the official fix notes; the only actionable conclusion for you is to measure your own machine.
Run a fixed disk-level and directory-size check once a month and write the numbers down. Anomalous growth versus last month is the earliest alarm — earlier than any news. This is this site's recommendation, not an official requirement.
Step-by-step SOP STEP BY STEP
- 01Read the waterline
Run
df -hand focus on the system disk's row: capacity and available space. If free space is clearly below where you remember it, keep going. - 02Size the Codex data directory
Run
du -sh ~/.codexfor the total footprint. There is no official 'normal' number — but measured against your own usage intensity, an abnormally large one is easy to spot. - 03Find the big hitters
Run
du -h -d 1 ~/.codex | sort -hto list first-level subdirectories sorted by size, and lock onto the abnormally swollen log or session directory. - 04Confirm, then clean
Spot-check modification times and file shapes inside the suspicious directory to confirm they're logs or session records — not configs or credentials. Delete with
rm -ri(interactive, item by item); never a blind recursive wipe. - 05Update the CLI
Upgrade Codex CLI to the latest version (install and upgrade commands per the official repo README) and check the release notes for whether a logging fix is on the list.
- 06Set up the patrol
Make
df -handdu -sh ~/.codexa monthly ritual and record the numbers each time. A sudden jump versus last month is the earliest alarm you'll get.
Keep reading NEXT
Want today's verdict? Back to the live radar.