zeroclaw-labs/zeroclaw · warning · DiagItem

low disk space: only {avail_mb} MB available

Error message

low disk space: only {avail_mb} MB available

What it means

A `zeroclaw doctor` warning from `check_workspace`: the filesystem holding `data_dir` reports fewer than 100 MB available (measured best-effort by shelling out to `df -m <data_dir>`). Low disk threatens session logs, state files, and probe writes before the OS returns ENOSPC. If `df` is unavailable the check is silently skipped, so seeing the warning means `df` ran and reported under the 100 MB threshold.

Source

Thrown at crates/zeroclaw-runtime/src/doctor/mod.rs:1511

            }
        }
        Err(e) => {
            items.push(DiagItem::error(
                cat,
                format!("directory is not writable: {e}"),
            ));
        }
    }

    // Disk space (best-effort via `df`)
    if let Some(avail_mb) = disk_available_mb(ws) {
        if avail_mb >= 100 {
            items.push(DiagItem::ok(
                cat,
                format!("disk space: {avail_mb} MB available"),
            ));
        } else {
            items.push(DiagItem::warn(
                cat,
                format!("low disk space: only {avail_mb} MB available"),
            ));
        }
    }

    let mut agent_aliases: Vec<&String> = config.agents.keys().collect();
    agent_aliases.sort();
    for alias in agent_aliases {
        let agent = config.agents.get(alias).expect("alias from keys()");
        if !agent.enabled {
            continue;
        }
        let agent_ws = config.agent_workspace_dir(alias);
        check_agent_file(&agent_ws, "SOUL.md", alias, cat, items);
        check_agent_file(&agent_ws, "AGENTS.md", alias, cat, items);
    }
}

View on GitHub (pinned to 88bb9c8533)

Solutions

1. Free space on the filesystem holding `data_dir` (clear logs/caches, remove old session data).
2. Or relocate `data_dir` in config to a volume with headroom.
3. Set up a disk-space alert/monitor so the daemon never reaches this state.
4. Re-run `zeroclaw doctor` and confirm it now reports `disk space: N MB available` as ok.

let out = std::process::Command::new("df").arg("-m").arg(&config.data_dir).output()?;
let avail_mb: u64 = String::from_utf8_lossy(&out.stdout)
    .lines().rev().find(|l| !l.trim().is_empty())
    .and_then(|l| l.split_whitespace().nth(3)?.parse().ok())
    .unwrap_or(0);
assert!(avail_mb >= 100, "data_dir filesystem has only {avail_mb} MB free");

// Best-effort, same posture as doctor: skip the check when df is unavailable
if let Some(avail_mb) = disk_available_mb(&config.data_dir) {
    if avail_mb < 100 { /* alert / refuse to start heavy jobs */ }
}

When it happens

Trigger: Running `zeroclaw doctor` when `df -m <data_dir>` reports < 100 MB available — typically a full root partition, a small tmpfs/container volume, or a quota-backed home directory.

Common situations: Long-running daemon accumulating session logs on a small VPS; container with a small writable layer; CI runner with a nearly full disk; `/tmp` mounted small with `data_dir` pointing into it.

Related errors

• [{alias}] {name} not found (optional)
• model route "{}" uses invalid model_provider "{}": {}
• embedding route "{}" uses invalid model_provider "{}": {}
• agent "{name}" uses invalid model_provider "{provider_ref}":
• OpenAI Codex credentials are signed in but no model provider

AI-assisted analysis of zeroclaw-labs/zeroclaw@88bb9c8533 (2026-08-23). Data as JSON: /api/errors/827e5fd7648b0940. Report an issue: GitHub.