vercel/turborepo · warning

{pad}• {base_msg}

Error message

{pad}• {base_msg}

What it means

Before a run, turbo resolves a RemoteCacheStatus and formats it via remote_cache_status_message (crates/turborepo-lib/src/run/mod.rs:161-213). Enabled/disabled statuses are printed as info; the Unavailable family (CouldNotConnect, AuthenticationFailed, UsageLimitExceeded, SpendingPaused, DisabledForTeam, UnexpectedServerError) is emitted as this warning with the reason inline, so builds proceed local-cache-only.

Source

Thrown at crates/turborepo-lib/src/run/mod.rs:455

                    self.filtered_pkgs.len()
                ),
            )
            .emit();
        }

        let (base_msg, is_warning) = remote_cache_status_message(
            self.remote_cache_status,
            &self.opts.api_client_opts.api_url,
        );

        let cache_status = if self.opts.run_opts.is_shared_worktree_cache {
            format!("{pad}• {base_msg}, using shared worktree cache")
        } else {
            format!("{pad}• {base_msg}")
        };

        if is_warning {
            turborepo_log::warn(
                turborepo_log::Source::turbo(turborepo_log::Subsystem::Run),
                cache_status,
            )
            .emit();
        } else {
            turborepo_log::info(
                turborepo_log::Source::turbo(turborepo_log::Subsystem::Run),
                cache_status,
            )
            .emit();
        }
        turborepo_log::info(
            turborepo_log::Source::turbo(turborepo_log::Subsystem::Run),
            "",
        )
        .emit();
    }

View on GitHub (pinned to f9245100cf)

Solutions

1. Match the parenthesized reason: 'Authentication failed' -> refresh credentials (turbo login or rotate TURBO_TOKEN/TURBO_TEAM); 'Could not connect' -> check TURBO_API/network/proxy; 'Usage limit exceeded'/'Spending paused' -> billing/plan; 'Disabled for this team' -> team settings.
2. Validate the token against the API endpoint manually (curl with the bearer token) to separate auth vs network issues.
3. For self-hosted caches, verify the endpoint's health and TLS before rerunning turbo.
4. Keep local caching on so builds degrade to local-only rather than fully cold.

Example fix

# before: stale CI token
TURBO_TOKEN=old-token TURBO_TEAM=acme turbo run build
# -> Remote caching unavailable (Authentication failed ...)

# after: rotate the secret, verify, rerun
curl -sS -H "Authorization: Bearer $TURBO_TOKEN" "$TURBO_API/v8/user" >/dev/null && echo ok
TURBO_TOKEN=new-token TURBO_TEAM=acme turbo run build

#!/usr/bin/env bash
# preflight remote cache before a long run
: "${TURBO_API:=https://api.vercel.com}"
if [ -n "$TURBO_TOKEN" ]; then
  code=$(curl -s -o /dev/null -w '%{http_code}' \
    -H "Authorization: Bearer $TURBO_TOKEN" "$TURBO_API/v8/user")
  [ "$code" = 200 ] || { echo "remote cache auth preflight failed: HTTP $code" >&2; exit 1; }
fi
turbo run build

When it happens

Trigger: Remote caching is configured (linked or TURBO_TOKEN present) but the preflight to the API fails: unreachable api url, invalid/expired token, team usage limit hit, spending paused, remote cache disabled for the team, or a 5xx from the server. is_warning=true only for Unavailable reasons.

Common situations: Expired TURBO_TOKEN in CI secrets, typo'd/stale TURBO_API, Vercel usage limits reached at month end, org spending pauses, or self-hosted cache endpoints being down. Builds succeed but every run is cold on remote.

Related errors

• no caches are enabled
• Remote cache is read-only, skipping upload
• Found both turbo.json and turbo.jsonc in the same directory:
• Found both turbo.json and turbo.jsonc in the same directory:
• refusing to write structured log to symlink: {}

AI-assisted analysis of vercel/turborepo@f9245100cf (2026-08-17). Data as JSON: /api/errors/a2037e11836ceaaa. Report an issue: GitHub.