fix(ci): make redis_setexz a safe no-op when redis is unavailable

[AztecBot]

Problem

The nightly barretenberg debug build failed immediately, before it could even request an EC2 instance (failing run):

--- Run barretenberg-debug CI ---

gzip: stdout: Broken pipe
##[error]Process completed with exit code 1.

Root cause

ci3/bootstrap_ec2 writes an initial log entry to redis at the very start of a run:

echo "CI booting..." | redis_setexz "$CI_LOG_ID" 300

redis_setexz was:

function redis_setexz {
  gzip | redis_cli -x SETEX $1 $2 &>/dev/null
}

When redis is unavailable, redis_cli (in ci3/source_redis) is a no-op that returns without reading its stdin. gzip then writes into a pipe whose read end is already closed and dies with SIGPIPE / "Broken pipe". Because ci3 runs under set -euo pipefail, that broken-pipe failure propagates out of the pipeline and kills the whole script before any instance is launched.

This is the direct cause of the observed failure: the run happened on the aztec-claude mirror, whose environment has no AWS creds and no BUILD_INSTANCE_SSH_KEY (all empty in the job env), so the redis tunnel is never opened and CI_REDIS_AVAILABLE=0. source_redis intentionally degrades gracefully in that case ("Log and test cache will be disabled") — but the very next redis_setexz call defeats that graceful degradation.

The same latent bug would take down the real nightly (or any CI run) any time the redis tunnel fails to establish, since bootstrap_ec2:24, cache_log's publish_log, and run_test_cmd all call redis_setexz unguarded, relying on it being a safe no-op. (denoise already works around it by guarding on CI_REDIS_AVAILABLE before calling it.)

Fix

Make redis_setexz itself a proper no-op when redis is unavailable, mirroring the existing redis_cli guard, and still drain stdin so the upstream pipeline producer doesn't get SIGPIPE:

function redis_setexz {
  if [ "$CI_REDIS_AVAILABLE" -ne 1 ]; then
    cat >/dev/null
    return 0
  fi
  gzip | redis_cli -x SETEX $1 $2 &>/dev/null
}

This fixes all unguarded callers at once and keeps the redis-available path unchanged.

Verification

Reproduced the failure under set -euo pipefail with CI_REDIS_AVAILABLE=0:

• Before: echo "CI booting..." | redis_setexz k 300 → pipeline dies with exit 141 (SIGPIPE) / "gzip: stdout: Broken pipe".
• After: same call returns 0 and the script continues; the redis-available path (stub redis_cli consuming stdin) also returns 0.

Notes

The scheduled nightly is guarded to only run on AztecProtocol/aztec-packages (barretenberg-nightly-debug-build.yml line 14); the failing run came from the aztec-claude mirror's older copy of that workflow that predates the guard, which is why it ran there at all. That mirror will self-correct on its next upstream sync — this PR fixes the underlying robustness bug that the mirror's credential-less environment exposed.

---

Created by claudebox · group: slackbot