Set -e + pipefail vs Manual Exit Checks: Which Guarantees Robust Deployment Log Analysis?

Answer First

For most deployment pipelines, set -e -o pipefail -o errtrace at the top of the script gives the strongest, most readable guarantee that a failure will abort the run. It catches every failing command in a pipeline, any command that exits non‑zero outside of exempted contexts, and propagates errors from functions. However, command substitutions and subshells that capture log output can still hide failures because the outer shell receives only the substitution’s output, not its exit status. In those cases, a single if [[ $? -ne 0 ]]; then …; fi after the substitution is the safest way to surface the error.

Why the Global Options Work

• -e exits on a non‑zero status unless the command is part of an if, while, until, &&, ||, or a conditional expression.
• -o pipefail makes the exit status of a pipeline the first non‑zero status, so a failure in the middle of cmd1 | cmd2 | cmd3 aborts the script.
• -o errtrace propagates trap ERR into functions and subshells, giving a single place to log the error.

When Manual Checks Are Still Needed

• Command substitution that must be inspected – e.g. LOG=$(grep -i "error" /var/log/app.log). set -e does not abort if grep fails; capture the status immediately: LOG=$(grep …); rc=$?; if [[ $rc -ne 0 ]]; then …; fi.
• Subshells that perform cleanup – a subshell spawned by (…) inherits -e, but the parent will continue unless you check the subshell’s exit code.
• Intentional non‑fatal failures – use || true only when you deliberately want to ignore a specific failure; otherwise, let -e surface it.

Practical Steps for Your Pipeline

1. #!/usr/bin/env bash
   set -e -o pipefail -o errtrace
   
   trap 'echo "ERROR: $BASH_COMMAND failed with exit $?'" >&2' ERR
   

2. Wrap any command substitution that might fail:

   LOG=$(grep -i "error" /var/log/app.log) || {
     echo "Log grep failed – aborting" >&2
     exit 1
   }
   

3. For subshells that need to report their status, capture it:

   RESULT=$( (some_long_command) ); rc=$?; if [[ $rc -ne 0 ]]; then echo "Subshell failed"; exit $rc; fi
   

4. Keep the rest of the script concise; the global options reduce boilerplate while still allowing targeted manual checks where the semantics of -e fall short.

Is a Missing Detail Important?

If your log‑parsing logic relies heavily on nested command substitutions or complex pipelines that produce intermediate files, let us know. That information can change whether a blanket set -e is sufficient or whether you need more granular checks.