fix(ci3): avoid redis log SIGPIPE when disabled

[AztecBot]

Summary

The nightly debug build failed before launching the EC2 build. The failing aztec-claude job ran ./.github/ci3.sh barretenberg-debug, which reached ci3/bootstrap_ec2 and immediately hit gzip: stdout: Broken pipe while initializing the dashboard log.

Root cause: redis_setexz always ran gzip | redis_cli .... When Redis is unavailable, redis_cli is intentionally a no-op and exits without reading stdin. Under set -o pipefail, gzip then exits on SIGPIPE and the whole CI script fails before printing the CI log URL or requesting the build instance.

This fixes redis_setexz so the disabled Redis path drains stdin and exits successfully, and keeps Redis upload failures non-fatal in the enabled path.

Tests

• ./ci3/tests/source_redis_test
• bash -n ci3/source_redis ci3/tests/source_redis_test
• ./ci3/bootstrap.sh was attempted; it passed the new test and then failed in existing ci3/tests/signal_test because Docker is unavailable in this container.

Refs https://github.com/AztecProtocol/aztec-claude/actions/runs/30146484544

---

Created by claudebox · group: slackbot