mirror of
https://github.com/boxpositron/absolute-dotfiles.git
synced 2026-10-24 08:23:19 +00:00
feat(docker-nuke): add script scaffold, arg parsing, and test harness
This commit is contained in:
David Ibia
Executable
+123
@@ -0,0 +1,123 @@
|
||||
#!/usr/bin/env bash
|
||||
# docker-nuke — capture diagnostics from a wedged Docker Desktop, then
|
||||
# force-kill and restart it.
|
||||
#
|
||||
# Written for bash 3.2 compatibility (macOS /bin/bash).
|
||||
|
||||
set -euo pipefail
|
||||
|
||||
# --- configuration ----------------------------------------------------------
|
||||
#
|
||||
# The SC2034 directives below are per-line on purpose, not file-level: each
|
||||
# constant is consumed by a later task, and its directive is deleted at that
|
||||
# point. A file-level disable would silence SC2034 for the whole file forever,
|
||||
# hiding a genuinely dead variable added later. By the final task none of these
|
||||
# directives should remain.
|
||||
|
||||
# shellcheck disable=SC2034
|
||||
HEALTH_TIMEOUT=6 # seconds to wait on `docker info` before calling it wedged
|
||||
# shellcheck disable=SC2034
|
||||
GRACEFUL_WAIT=8 # seconds to let Docker quit politely
|
||||
# shellcheck disable=SC2034
|
||||
TERM_WAIT=5 # seconds to wait after SIGTERM before SIGKILL
|
||||
# shellcheck disable=SC2034
|
||||
RESTART_TIMEOUT=90 # seconds to wait for the daemon to come back
|
||||
# shellcheck disable=SC2034
|
||||
POLL_INTERVAL=2 # seconds between daemon health polls
|
||||
# shellcheck disable=SC2034
|
||||
SAMPLE_DURATION=2 # seconds of stack sampling per process
|
||||
# shellcheck disable=SC2034
|
||||
SAMPLE_TIMEOUT=10 # hard cap on a single `sample` invocation
|
||||
# shellcheck disable=SC2034
|
||||
DUMP_RETENTION=10 # how many dump directories to keep
|
||||
# shellcheck disable=SC2034
|
||||
LOG_TAIL_LINES=2000 # lines of Docker log to copy (keeps dumps small)
|
||||
|
||||
# shellcheck disable=SC2034
|
||||
DOCKER_APP="/Applications/Docker.app"
|
||||
# shellcheck disable=SC2034
|
||||
DOCKER_LOG_DIR="$HOME/Library/Containers/com.docker.docker/Data/log"
|
||||
# shellcheck disable=SC2034
|
||||
DOCKER_RAW="$HOME/Library/Containers/com.docker.docker/Data/vms/0/data/Docker.raw"
|
||||
# shellcheck disable=SC2034
|
||||
CRASH_DIR="$HOME/Library/Logs/DiagnosticReports"
|
||||
|
||||
# Overridable so tests never touch real state.
|
||||
# shellcheck disable=SC2034
|
||||
DUMP_ROOT="${DOCKER_NUKE_DUMP_ROOT:-$HOME/.local/state/docker-nuke}"
|
||||
|
||||
OPT_FORCE=0
|
||||
OPT_NO_RESTART=0
|
||||
OPT_DRY_RUN=0
|
||||
OPT_DEEP=0
|
||||
|
||||
# --- helpers ----------------------------------------------------------------
|
||||
|
||||
bin_exists() { command -v "$1" >/dev/null 2>&1; }
|
||||
|
||||
log() { printf '%s\n' "$*"; }
|
||||
warn() { printf '%s\n' "$*" >&2; }
|
||||
|
||||
usage() {
|
||||
cat <<'EOF'
|
||||
docker-nuke: force-restart a wedged Docker Desktop, capturing diagnostics first
|
||||
|
||||
Usage:
|
||||
docker-nuke [options]
|
||||
|
||||
Options:
|
||||
-f, --force Skip the confirmation shown when Docker looks healthy
|
||||
-n, --no-restart Kill Docker but do not relaunch it
|
||||
--dry-run Report what would be killed and dumped; change nothing
|
||||
--deep Also stop root launchd helpers (com.docker.vmnetd); needs sudo
|
||||
-h, --help Show this help
|
||||
|
||||
Diagnostics are written to ~/.local/state/docker-nuke/<timestamp>/ before
|
||||
anything is killed. The most recent 10 dumps are kept.
|
||||
|
||||
Exit codes:
|
||||
0 Docker killed and confirmed responsive again
|
||||
1 Daemon did not come back within the restart timeout
|
||||
2 Usage error, or aborted at the healthy-daemon prompt
|
||||
EOF
|
||||
}
|
||||
|
||||
parse_args() {
|
||||
OPT_FORCE=0
|
||||
OPT_NO_RESTART=0
|
||||
OPT_DRY_RUN=0
|
||||
OPT_DEEP=0
|
||||
while [ "$#" -gt 0 ]; do
|
||||
# shellcheck disable=SC2034 # OPT_* are read by main once it is wired up
|
||||
case "$1" in
|
||||
-f|--force) OPT_FORCE=1 ;;
|
||||
-n|--no-restart) OPT_NO_RESTART=1 ;;
|
||||
--dry-run) OPT_DRY_RUN=1 ;;
|
||||
--deep) OPT_DEEP=1 ;;
|
||||
-h|--help) usage; return 3 ;;
|
||||
*)
|
||||
warn "docker-nuke: unknown option '$1'"
|
||||
usage >&2
|
||||
return 2
|
||||
;;
|
||||
esac
|
||||
shift
|
||||
done
|
||||
return 0
|
||||
}
|
||||
|
||||
# --- entrypoint -------------------------------------------------------------
|
||||
|
||||
main() {
|
||||
local rc=0
|
||||
parse_args "$@" || rc=$?
|
||||
if [ "$rc" -eq 3 ]; then return 0; fi
|
||||
if [ "$rc" -ne 0 ]; then return "$rc"; fi
|
||||
log "docker-nuke: not yet implemented"
|
||||
return 0
|
||||
}
|
||||
|
||||
# Only run when executed, not when sourced by the test harness.
|
||||
if [ "${BASH_SOURCE[0]}" = "${0}" ]; then
|
||||
main "$@"
|
||||
fi
|
||||
Executable
+99
@@ -0,0 +1,99 @@
|
||||
#!/usr/bin/env bash
|
||||
# Test harness for docker-nuke. No external dependencies.
|
||||
# Run: ./test-docker-nuke.sh
|
||||
|
||||
set -uo pipefail # deliberately not -e: a failing assertion must not abort the run
|
||||
|
||||
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
|
||||
TESTS_RUN=0
|
||||
TESTS_FAILED=0
|
||||
TMPROOT=""
|
||||
|
||||
setup_tmp() {
|
||||
TMPROOT="$(mktemp -d)"
|
||||
export DOCKER_NUKE_DUMP_ROOT="$TMPROOT/dumps"
|
||||
}
|
||||
|
||||
teardown_tmp() {
|
||||
if [ -n "$TMPROOT" ] && [ -d "$TMPROOT" ]; then
|
||||
rm -rf "$TMPROOT"
|
||||
fi
|
||||
TMPROOT=""
|
||||
}
|
||||
|
||||
pass() {
|
||||
TESTS_RUN=$((TESTS_RUN + 1))
|
||||
printf ' ok %s\n' "$1"
|
||||
}
|
||||
|
||||
fail() {
|
||||
TESTS_RUN=$((TESTS_RUN + 1))
|
||||
TESTS_FAILED=$((TESTS_FAILED + 1))
|
||||
printf ' FAIL %s\n %s\n' "$1" "$2"
|
||||
}
|
||||
|
||||
assert_eq() {
|
||||
local desc="$1" expected="$2" actual="$3"
|
||||
if [ "$expected" = "$actual" ]; then
|
||||
pass "$desc"
|
||||
else
|
||||
fail "$desc" "expected [$expected] got [$actual]"
|
||||
fi
|
||||
}
|
||||
|
||||
assert_contains() {
|
||||
local desc="$1" haystack="$2" needle="$3"
|
||||
case "$haystack" in
|
||||
*"$needle"*) pass "$desc" ;;
|
||||
*) fail "$desc" "expected to contain [$needle] got [$haystack]" ;;
|
||||
esac
|
||||
}
|
||||
|
||||
assert_not_contains() {
|
||||
local desc="$1" haystack="$2" needle="$3"
|
||||
case "$haystack" in
|
||||
*"$needle"*) fail "$desc" "expected NOT to contain [$needle] got [$haystack]" ;;
|
||||
*) pass "$desc" ;;
|
||||
esac
|
||||
}
|
||||
|
||||
# Source the script under test. Its trailing BASH_SOURCE/$0 guard means main
|
||||
# runs only when the script is executed directly, so sourcing it here defines
|
||||
# the functions without triggering a real nuke.
|
||||
setup_tmp
|
||||
# shellcheck source=/dev/null
|
||||
source "$SCRIPT_DIR/docker-nuke"
|
||||
set +e # the sourced script sets -e; undo it for the harness
|
||||
|
||||
printf '\n=== argument parsing ===\n'
|
||||
|
||||
test_parse_args() {
|
||||
parse_args --force
|
||||
assert_eq "--force sets OPT_FORCE" "1" "$OPT_FORCE"
|
||||
|
||||
parse_args -n
|
||||
assert_eq "-n sets OPT_NO_RESTART" "1" "$OPT_NO_RESTART"
|
||||
|
||||
parse_args --dry-run
|
||||
assert_eq "--dry-run sets OPT_DRY_RUN" "1" "$OPT_DRY_RUN"
|
||||
|
||||
parse_args --deep
|
||||
assert_eq "--deep sets OPT_DEEP" "1" "$OPT_DEEP"
|
||||
|
||||
parse_args
|
||||
assert_eq "no flags leaves OPT_FORCE off" "0" "$OPT_FORCE"
|
||||
|
||||
parse_args --bogus >/dev/null 2>&1
|
||||
assert_eq "unknown flag returns 2" "2" "$?"
|
||||
|
||||
local help_text
|
||||
help_text="$(usage)"
|
||||
assert_contains "usage mentions --dry-run" "$help_text" "--dry-run"
|
||||
assert_contains "usage mentions --deep" "$help_text" "--deep"
|
||||
}
|
||||
test_parse_args
|
||||
|
||||
printf '\n=== summary ===\n'
|
||||
teardown_tmp
|
||||
printf 'ran %d, failed %d\n' "$TESTS_RUN" "$TESTS_FAILED"
|
||||
[ "$TESTS_FAILED" -eq 0 ]
|
||||
Reference in New Issue
Block a user