Remove Snapper timeline snapshots leaked by earlier defaults
Installs from before the Snapper setup was normalized ran hourly timeline snapshots. Newer configs stopped creating them but never deleted the existing ones, and number cleanup skips snapshots marked Cleanup=timeline, so they sit there forever: one machine installed from the 2026-05-11 ISO had accumulated 592 of them, silently pinning 219 GB of disk. The limine-snapper-sync limit-mismatch warning that would have surfaced this is disabled by default since the notifier migration. Delete leaked timeline snapshots in batches of 20 (a single mass delete can die on a DBus timeout partway through), and only when TIMELINE_CREATE="no" so anyone who deliberately re-enabled timeline snapshotting keeps their setup untouched. The drain is best effort: a batch that fails is skipped rather than aborting the migration run and everything queued behind it, since the next run re-lists whatever is left.
This commit is contained in:
Martin Bastien
@@ -0,0 +1,43 @@
|
||||
echo "Remove Snapper timeline snapshots leaked by earlier defaults"
|
||||
|
||||
SNAPPER_CONFIG_PATH="${OMARCHY_SNAPPER_CONFIG_PATH:-/etc/snapper/configs/root}"
|
||||
|
||||
as_root() {
|
||||
if (( EUID == 0 )); then
|
||||
"$@"
|
||||
else
|
||||
sudo "$@"
|
||||
fi
|
||||
}
|
||||
|
||||
command -v snapper >/dev/null || exit 0
|
||||
[[ -f $SNAPPER_CONFIG_PATH ]] || exit 0
|
||||
|
||||
# Only clean up when timeline snapshotting is off, as Omarchy configures it.
|
||||
# Anyone who deliberately turned it back on keeps their snapshots.
|
||||
grep -qFx 'TIMELINE_CREATE="no"' "$SNAPPER_CONFIG_PATH" || exit 0
|
||||
|
||||
# Earlier installs ran hourly timeline snapshots. Later configs stopped
|
||||
# creating them but never deleted the existing ones, and number cleanup
|
||||
# skips snapshots marked Cleanup=timeline, so they pile up forever:
|
||||
# hundreds of snapshots pinning 100+ GB of extents on long-running machines.
|
||||
leaked=$(as_root snapper -c root --csvout list --columns number,cleanup 2>/dev/null | awk -F, '$2 == "timeline" { print $1 }' || true)
|
||||
[[ -n $leaked ]] || exit 0
|
||||
|
||||
echo "Deleting $(wc -w <<<"$leaked") leaked timeline snapshots (disk space is reclaimed in the background)"
|
||||
|
||||
# Delete in small batches; one big delete can die on a DBus timeout partway.
|
||||
# A failed batch must not take the rest of the migration run down with it, so
|
||||
# the drain is best effort: whatever survives is picked up by the next run.
|
||||
batch=()
|
||||
for number in $leaked; do
|
||||
batch+=("$number")
|
||||
if (( ${#batch[@]} == 20 )); then
|
||||
as_root snapper -c root delete "${batch[@]}" || true
|
||||
batch=()
|
||||
fi
|
||||
done
|
||||
|
||||
if (( ${#batch[@]} > 0 )); then
|
||||
as_root snapper -c root delete "${batch[@]}" || true
|
||||
fi
|
||||
@@ -0,0 +1,104 @@
|
||||
#!/bin/bash
|
||||
|
||||
set -euo pipefail
|
||||
|
||||
source "$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)/base-test.sh"
|
||||
|
||||
leak_migration=$(grep -rl 'timeline snapshots leaked by earlier defaults' "$ROOT/migrations" | head -n 1 || true)
|
||||
[[ -n $leak_migration ]] || fail "Snapper timeline leak migration exists"
|
||||
|
||||
test_tmp=$(mktemp -d)
|
||||
trap 'rm -rf "$test_tmp"' EXIT
|
||||
|
||||
fake_bin="$test_tmp/bin"
|
||||
mkdir -p "$fake_bin"
|
||||
|
||||
cat >"$fake_bin/sudo" <<'STUB'
|
||||
#!/bin/bash
|
||||
exec "$@"
|
||||
STUB
|
||||
chmod +x "$fake_bin/sudo"
|
||||
|
||||
cat >"$fake_bin/snapper" <<'STUB'
|
||||
#!/bin/bash
|
||||
printf 'snapper %s\n' "$*" >>"$TEST_LOG"
|
||||
if [[ "$*" == *"--csvout list"* ]]; then
|
||||
echo "number,cleanup"
|
||||
for i in $(seq 1 45); do
|
||||
echo "$i,timeline"
|
||||
done
|
||||
echo "100,number"
|
||||
echo "101,"
|
||||
fi
|
||||
STUB
|
||||
chmod +x "$fake_bin/snapper"
|
||||
|
||||
snapper_config="$test_tmp/root"
|
||||
printf '%s\n' 'TIMELINE_CREATE="no"' 'NUMBER_CLEANUP="yes"' >"$snapper_config"
|
||||
|
||||
TEST_LOG="$test_tmp/calls.log" \
|
||||
PATH="$fake_bin:$PATH" \
|
||||
OMARCHY_SNAPPER_CONFIG_PATH="$snapper_config" \
|
||||
bash -euo pipefail "$leak_migration" >/dev/null
|
||||
|
||||
deletes=$(grep -c '^snapper -c root delete ' "$test_tmp/calls.log" || true)
|
||||
[[ $deletes -eq 3 ]] || fail "leak migration deletes snapshots in batches" "expected 3 delete calls, got $deletes"
|
||||
|
||||
first_batch=$(grep -m1 '^snapper -c root delete ' "$test_tmp/calls.log")
|
||||
[[ $first_batch == "snapper -c root delete $(seq -s ' ' 1 20)" ]] || fail "leak migration caps delete batches at 20 snapshots" "$first_batch"
|
||||
|
||||
last_batch=$(grep '^snapper -c root delete ' "$test_tmp/calls.log" | tail -n 1)
|
||||
[[ $last_batch == "snapper -c root delete $(seq -s ' ' 41 45)" ]] || fail "leak migration deletes the final partial batch" "$last_batch"
|
||||
|
||||
! grep -E '^snapper -c root delete .*\b(100|101)\b' "$test_tmp/calls.log" || fail "leak migration only deletes timeline snapshots"
|
||||
pass "leak migration removes leaked timeline snapshots in batches and keeps the rest"
|
||||
|
||||
# omarchy-migrate runs under set -e, so a batch that dies on a DBus timeout
|
||||
# would otherwise abort the run and skip every migration queued behind it.
|
||||
: >"$test_tmp/calls.log"
|
||||
printf '%s\n' 'TIMELINE_CREATE="no"' 'NUMBER_CLEANUP="yes"' >"$snapper_config"
|
||||
|
||||
cat >"$fake_bin/snapper" <<'STUB'
|
||||
#!/bin/bash
|
||||
printf 'snapper %s\n' "$*" >>"$TEST_LOG"
|
||||
if [[ "$*" == *"--csvout list"* ]]; then
|
||||
echo "number,cleanup"
|
||||
for i in $(seq 1 45); do
|
||||
echo "$i,timeline"
|
||||
done
|
||||
exit 0
|
||||
fi
|
||||
echo "failure: dbus timeout" >&2
|
||||
exit 1
|
||||
STUB
|
||||
|
||||
TEST_LOG="$test_tmp/calls.log" \
|
||||
PATH="$fake_bin:$PATH" \
|
||||
OMARCHY_SNAPPER_CONFIG_PATH="$snapper_config" \
|
||||
bash -euo pipefail "$leak_migration" >/dev/null 2>&1 ||
|
||||
fail "leak migration survives a failed delete batch"
|
||||
|
||||
deletes=$(grep -c '^snapper -c root delete ' "$test_tmp/calls.log" || true)
|
||||
[[ $deletes -eq 3 ]] || fail "leak migration keeps draining after a failed batch" "expected 3 delete calls, got $deletes"
|
||||
pass "leak migration tolerates a batch that fails partway"
|
||||
|
||||
: >"$test_tmp/calls.log"
|
||||
printf '%s\n' 'TIMELINE_CREATE="yes"' >"$snapper_config"
|
||||
|
||||
TEST_LOG="$test_tmp/calls.log" \
|
||||
PATH="$fake_bin:$PATH" \
|
||||
OMARCHY_SNAPPER_CONFIG_PATH="$snapper_config" \
|
||||
bash -euo pipefail "$leak_migration" >/dev/null
|
||||
|
||||
[[ ! -s $test_tmp/calls.log ]] || fail "leak migration leaves deliberate timeline setups alone"
|
||||
pass "leak migration skips systems where timeline snapshots are intentional"
|
||||
|
||||
: >"$test_tmp/calls.log"
|
||||
|
||||
TEST_LOG="$test_tmp/calls.log" \
|
||||
PATH="$fake_bin:$PATH" \
|
||||
OMARCHY_SNAPPER_CONFIG_PATH="$test_tmp/missing" \
|
||||
bash -euo pipefail "$leak_migration" >/dev/null
|
||||
|
||||
[[ ! -s $test_tmp/calls.log ]] || fail "leak migration skips systems without a Snapper root config"
|
||||
pass "leak migration is a no-op without Snapper configured"
|
||||
Reference in New Issue
Block a user