Files
ihasmail/deploy.example.sh
T
jcoffey-dev 82108ebd97 Stop the deploy script rewriting itself mid-run
Moving the deploy script into the repo put it inside the checkout it
resets, and bash does not read a script all at once -- it reads as it
goes, by byte offset. `git reset --hard` replacing the file underneath a
running shell makes it stop wherever it had reached.

Silently, and with exit status 0. A three-line demonstration:

    echo "line A"
    cat > "$0" <<'NEW'
    echo "REWRITTEN"
    NEW
    echo "line B"

prints "line A" and nothing else, and exits 0. In a deploy that means
building the image, then stopping before the container is replaced, and
reporting success -- so the old container keeps serving while everything
says the new one shipped.

It only bites when a deploy carries a change to the deploy script
itself, which is rare enough to be baffling when it happens and exactly
the sort of quiet failure this project keeps paying for.

The script now re-execs from a copy outside the tree before touching
git, so the file being run cannot change while it runs, and removes the
copy on exit. Running it from outside the checkout -- as the host did
before this moved into the repo -- skips all of that.

The trap is an `if` rather than `[ -n ... ] && trap`, which would leave
the not-re-exec'd path resting on errexit ignoring a failed left operand
of &&. It does ignore it, but a deploy script is a poor place to depend
on knowing that.

Verified: with the guard, a script that overwrites itself mid-run
completes every later line and cleans up its copy; without it, the lines
after the rewrite never run.
2026-08-26 10:26:58 -07:00

187 lines
7.5 KiB
Bash
Executable File

#!/bin/bash
# Redeploy ihasmail on a single-host Docker setup, from a git checkout.
#
# Copy it, or run it as-is and set the variables below in the environment.
# Nothing here is specific to any one host: the defaults describe the shape of
# a deployment rather than anyone's particular one.
#
# Usage: ./deploy.sh [git-ref] [-y|--yes] [-n|--dry-run]
#
# Three guards stand between a careless run and production:
#
# .deploy-hold commits that must not reach prod yet, one per line. If the
# target contains one that is not already deployed, the deploy
# is refused outright -- `--yes` does not override it. Clearing
# a hold means deleting its line, which is a deliberate edit.
#
# confirmation anything introducing new commits is listed first and has to
# be confirmed. Over SSH, where there is no terminal to answer
# on, that means passing --yes: a bare `deploy.sh` cannot ship
# whatever main happens to have picked up since the last
# release.
#
# --dry-run runs both guards, says what it would deploy, and stops before
# building or touching the container.
#
# The container is replaced rather than restarted, because the image is rebuilt
# from the new checkout. Data lives in a named volume and survives that; the
# environment file is never read here, only handed to Docker.
set -euo pipefail
# --- what to deploy, and where ----------------------------------------------
# The checkout to deploy from. It must be a git clone: the version number is
# read from its history (see scripts/version.mjs).
APP="${IHASMAIL_APP:-$HOME/apps/ihasmail}"
# Environment file passed to the container. Keep it outside the repo's tracked
# files -- it holds APP_SECRET and the upstream URL. Never read by this script.
ENVF="${IHASMAIL_ENV:-$APP/.env.production}"
# Commits held back from production, one per line; blank or missing is fine.
HOLD="${IHASMAIL_HOLD:-$APP/.deploy-hold}"
# Container name, and where to publish it. The default binds to loopback only,
# for a reverse proxy in front (see Caddyfile.example / nginx.example.conf).
NAME="${IHASMAIL_NAME:-ihasmail}"
BIND="${IHASMAIL_BIND:-127.0.0.1:8090}"
# Named volume for /data (sessions).
VOLUME="${IHASMAIL_VOLUME:-ihasmail-data}"
# Image repository. Each build is tagged with its version as well, so an
# earlier one can be run again without rebuilding it.
IMAGE_REPO="${IHASMAIL_IMAGE:-ihasmail}"
# How long to wait for the new container to report healthy, in seconds.
HEALTH_TIMEOUT="${IHASMAIL_HEALTH_TIMEOUT:-30}"
# --- run from a copy, if this script lives in the checkout it resets ---------
# `git reset --hard` below rewrites the working tree, and this script may be
# part of it. Bash does not read a script all at once -- it reads as it goes,
# by byte offset -- so a file replaced underneath it makes the shell stop
# wherever it had reached. Silently, and with exit status 0: a deploy that
# stopped halfway would report success. Re-exec from a copy outside the tree so
# the file being run cannot change while it runs.
SELF="$(readlink -f "$0")"
APP_REAL="$(readlink -f "$APP" 2>/dev/null || printf '%s' "$APP")"
if [ -z "${IHASMAIL_REEXEC:-}" ] && [ "${SELF#"$APP_REAL"/}" != "$SELF" ]; then
COPY="$(mktemp "${TMPDIR:-/tmp}/ihasmail-deploy.XXXXXX")"
cat "$SELF" > "$COPY"
chmod +x "$COPY"
IHASMAIL_REEXEC=1 exec "$COPY" "$@"
fi
# The copy has served its purpose once we exit; the shell has finished reading
# it by then.
if [ -n "${IHASMAIL_REEXEC:-}" ]; then
trap 'rm -f "$SELF"' EXIT
fi
REF=""
ASSUME_YES=0
DRY_RUN=0
for arg in "$@"; do
case "$arg" in
-y|--yes) ASSUME_YES=1 ;;
-n|--dry-run) DRY_RUN=1 ;;
-h|--help) sed -n '2,28p' "$0"; exit 0 ;;
-*) echo "unknown option: $arg" >&2; exit 2 ;;
*)
if [ -n "$REF" ]; then echo "give at most one git-ref (got '$REF' and '$arg')" >&2; exit 2; fi
REF="$arg" ;;
esac
done
REF="${REF:-origin/main}"
cd "$APP"
git fetch --quiet origin
if ! TARGET=$(git rev-parse --verify --quiet "${REF}^{commit}"); then
echo "!! no such commit: $REF" >&2
exit 2
fi
CURRENT=$(git rev-parse --verify HEAD)
# --- guard 1: commits held back from production -----------------------------
if [ -f "$HOLD" ]; then
blocked=""
while IFS= read -r line || [ -n "$line" ]; do
line="${line%%#*}"
line="$(printf '%s' "$line" | tr -d '[:space:]')"
[ -z "$line" ] && continue
if ! held=$(git rev-parse --verify --quiet "${line}^{commit}"); then
echo " (hold list names '$line', which this checkout does not know -- ignoring)" >&2
continue
fi
# Only a problem if the target carries it and production does not already.
if git merge-base --is-ancestor "$held" "$TARGET" && ! git merge-base --is-ancestor "$held" "$CURRENT"; then
blocked="${blocked} $(git log --oneline -1 "$held")"$'\n'
fi
done < "$HOLD"
if [ -n "$blocked" ]; then
echo "!! refusing to deploy $REF: it contains commits held back from production:" >&2
printf '%s' "$blocked" >&2
echo " listed in $HOLD -- delete the line to clear the hold, or deploy a ref without it." >&2
exit 1
fi
fi
# --- guard 2: say what is being introduced, and get a yes --------------------
NEW=$(git log --oneline "$CURRENT..$TARGET")
if [ -n "$NEW" ]; then
echo "==> $(git log --oneline -1 "$CURRENT") -> $(git log --oneline -1 "$TARGET")"
echo "==> introduces:"
printf '%s\n' "$NEW" | sed 's/^/ /'
if [ "$ASSUME_YES" -ne 1 ]; then
if [ -t 0 ]; then
read -r -p "deploy these to production? [y/N] " reply
case "$reply" in
y|Y|yes|YES) ;;
*) echo "aborted."; exit 1 ;;
esac
else
echo "!! refusing: this introduces new commits and there is no terminal to confirm on." >&2
echo " re-run with --yes if that is what you mean, or name the ref you want." >&2
exit 1
fi
fi
else
echo "==> already at $(git log --oneline -1 "$TARGET"); rebuilding"
fi
if [ "$DRY_RUN" -eq 1 ]; then
echo "==> dry run: would deploy $(git log --oneline -1 "$TARGET"); nothing was changed"
exit 0
fi
git reset --hard --quiet "$TARGET"
# The version is worked out here, from the checkout, because the image build
# cannot: .dockerignore keeps .git out of the build context. Without this the
# build falls back to the base version in package.json and every deployment
# reports the same number -- see "Version numbers" in the README.
VERSION="$(node scripts/version.mjs)"
# A Docker tag may not contain "+", which a version for a commit that did not
# come through a pull request does: 2.16.57+g1fa6578. The image is tagged with
# the "+" turned into "-"; what the build is *told* it is keeps the real form,
# so About and /api/health still report it correctly.
TAG="${VERSION//+/-}"
echo "==> building $(git log --oneline -1) as v$VERSION"
docker build \
--build-arg IHASMAIL_VERSION="$VERSION" \
-t "$IMAGE_REPO:$TAG" \
-t "$IMAGE_REPO:current" \
.
echo "==> restarting container"
docker rm -f "$NAME" >/dev/null 2>&1 || true
docker run -d --name "$NAME" --restart unless-stopped \
-p "$BIND:8080" --env-file "$ENVF" -v "$VOLUME:/data" "$IMAGE_REPO:$TAG" >/dev/null
for _ in $(seq 1 "$HEALTH_TIMEOUT"); do
if health=$(curl -sf "http://$BIND/api/health"); then
echo "==> healthy: $health"
exit 0
fi
sleep 1
done
echo "!! did not become healthy after ${HEALTH_TIMEOUT}s; logs:" >&2
docker logs "$NAME" 2>&1 | tail -20 >&2
echo "!! the previous image is still tagged, if you need it back:" >&2
docker images "$IMAGE_REPO" --format ' {{.Repository}}:{{.Tag}} {{.CreatedSince}}' | head -5 >&2
exit 1