feat(cache): purge stale cargo cache entries in retention run
TODO 2.9 recorded a retention policy that warned about low disk while freeing 0.0 GB twice in a row, because it governed artifacts, logs and IP leases -- 52 KB in total on this host -- while 195 G of backups sat outside its reach. Adding an unbounded directory of ~645 MB entries, one per Cargo.lock change, without extending retention in the same change would repeat that at a larger scale. The window is the tighter of the global retention and the cache's own, so the aggressive low-disk mode reaches the cache rather than stopping at the light directories. Sizes are read before deleting, not after. Summing st_size over files that have already been unlinked yields 0 -- which is exactly how a policy ends up reporting that it freed nothing while doing real work. A test asserts the reported figure.
This commit is contained in:
@@ -14,7 +14,8 @@ from pathlib import Path
|
||||
|
||||
import click
|
||||
|
||||
from ci_orchestrator.config import load_config
|
||||
from ci_orchestrator.cargo_cache import CargoCacheStore
|
||||
from ci_orchestrator.config import Config, load_config
|
||||
|
||||
|
||||
@click.group()
|
||||
@@ -93,6 +94,7 @@ def retention_run(
|
||||
|
||||
_purge_old_dirs(art_path, "artifact", cutoff, what_if=what_if)
|
||||
_purge_old_dirs(log_path, "log", cutoff, what_if=what_if)
|
||||
_purge_cargo_cache(config, effective_days, what_if=what_if)
|
||||
|
||||
usage_after = shutil.disk_usage(str(check_path))
|
||||
free_gb_after = usage_after.free / (1024**3)
|
||||
@@ -132,3 +134,44 @@ def _purge_old_dirs(base: Path, label: str, cutoff: datetime, *, what_if: bool)
|
||||
else:
|
||||
shutil.rmtree(d, ignore_errors=True)
|
||||
click.echo(f"[RetentionPolicy] Purged {label} ({age_days}d): {d.name}")
|
||||
|
||||
|
||||
def _purge_cargo_cache(config: Config, effective_days: int, *, what_if: bool) -> None:
|
||||
"""Purge stale cargo cache entries.
|
||||
|
||||
Entries are the heaviest thing this policy governs -- hundreds of megabytes
|
||||
each, one per Cargo.lock change -- so leaving them out would repeat TODO
|
||||
2.9, where retention warned about low disk while freeing 0.0 GB because
|
||||
everything that actually weighed was outside its reach.
|
||||
|
||||
The window is the tighter of the global retention and the cache's own, so
|
||||
the aggressive low-disk mode reaches the cache too.
|
||||
"""
|
||||
if not config.cargo_cache.enabled:
|
||||
click.echo("[RetentionPolicy] cargo cache: disabled — skipping.")
|
||||
return
|
||||
|
||||
days = min(effective_days, config.cargo_cache.retention_days)
|
||||
cutoff = datetime.now(tz=UTC) - timedelta(days=days)
|
||||
store = CargoCacheStore(config.cargo_cache.root)
|
||||
|
||||
# Size the victims first: after a real prune the files are gone and their
|
||||
# bytes are unreadable, which is how a policy ends up reporting 0.0 GB
|
||||
# freed while doing real work (TODO 2.9).
|
||||
doomed = store.prune_older_than(cutoff, what_if=True)
|
||||
if not doomed:
|
||||
click.echo(
|
||||
f"[RetentionPolicy] cargo cache: nothing to purge "
|
||||
f"(cutoff: {cutoff.strftime('%Y-%m-%d')})."
|
||||
)
|
||||
return
|
||||
freed_mb = sum(p.stat().st_size for p in doomed if p.exists()) / 1_048_576
|
||||
|
||||
if not what_if:
|
||||
store.prune_older_than(cutoff)
|
||||
|
||||
verb = "WhatIf: would purge" if what_if else "Purged"
|
||||
click.echo(
|
||||
f"[RetentionPolicy] {verb} {len(doomed)} cargo cache "
|
||||
f"{'entry' if len(doomed) == 1 else 'entries'} ({freed_mb:.0f} MB)"
|
||||
)
|
||||
|
||||
@@ -11,6 +11,7 @@ from click.testing import CliRunner
|
||||
|
||||
import ci_orchestrator.commands.retention as ret
|
||||
from ci_orchestrator.__main__ import cli
|
||||
from ci_orchestrator.config import Config
|
||||
|
||||
# ── helpers ───────────────────────────────────────────────────────────────────
|
||||
|
||||
@@ -191,3 +192,134 @@ def test_purge_old_log_dirs(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) ->
|
||||
assert result.exit_code == 0, result.output
|
||||
assert not old.exists()
|
||||
assert "Purged log" in result.output
|
||||
|
||||
|
||||
# ── cargo cache entries ───────────────────────────────────────────────────────
|
||||
|
||||
|
||||
def _cache_config(
|
||||
tmp_path: Path, *, enabled: bool = True, days: int = 30
|
||||
) -> tuple[Path, Config]:
|
||||
"""Point the loaded config at a throwaway cache store.
|
||||
|
||||
Returns ``(store root, config)`` — the root so tests can seed entries, the
|
||||
config so they can substitute it for ``load_config``.
|
||||
"""
|
||||
from ci_orchestrator.config import BackendConfig, CargoCacheConfig, Config, Paths
|
||||
|
||||
root = tmp_path / "cache" / "cargo"
|
||||
cfg = Config(
|
||||
paths=Paths(
|
||||
root=tmp_path,
|
||||
templates=tmp_path / "t",
|
||||
build_vms=tmp_path / "b",
|
||||
artifacts=tmp_path / "artifacts",
|
||||
keys=tmp_path / "k",
|
||||
),
|
||||
backend=BackendConfig(),
|
||||
cargo_cache=CargoCacheConfig(enabled=enabled, root=root, retention_days=days),
|
||||
)
|
||||
return root, cfg
|
||||
|
||||
|
||||
def _seed_entry(root: Path, digest: str, *, age_days: int, size: int = 1024) -> Path:
|
||||
import os
|
||||
|
||||
from ci_orchestrator.cargo_cache import CargoCacheStore, cache_key
|
||||
|
||||
store = CargoCacheStore(root)
|
||||
src = root.parent / f"seed-{digest}.tar.zst"
|
||||
src.parent.mkdir(parents=True, exist_ok=True)
|
||||
src.write_bytes(b"x" * size)
|
||||
entry = store.publish(cache_key(lock_digest=digest, rustc_version="r", host_triple="t"), src)
|
||||
stamp = time.time() - age_days * 86400
|
||||
os.utime(entry, (stamp, stamp))
|
||||
return entry
|
||||
|
||||
|
||||
def test_retention_purges_stale_cargo_cache_entries(
|
||||
tmp_path: Path, monkeypatch: pytest.MonkeyPatch
|
||||
) -> None:
|
||||
root, cfg = _cache_config(tmp_path)
|
||||
old = _seed_entry(root, "old", age_days=90)
|
||||
fresh = _seed_entry(root, "new", age_days=1)
|
||||
|
||||
monkeypatch.setattr(ret, "load_config", lambda: cfg)
|
||||
monkeypatch.setattr(ret.shutil, "disk_usage", lambda _p: _fake_usage(free_gb=200))
|
||||
|
||||
result = CliRunner().invoke(cli, ["retention", "run"])
|
||||
assert result.exit_code == 0, result.output
|
||||
assert not old.exists(), "stale entry should be purged"
|
||||
assert fresh.exists(), "fresh entry should survive"
|
||||
|
||||
|
||||
def test_retention_reports_the_bytes_it_freed(
|
||||
tmp_path: Path, monkeypatch: pytest.MonkeyPatch
|
||||
) -> None:
|
||||
"""TODO 2.9: a policy that frees space without saying how much is the bug."""
|
||||
root, cfg = _cache_config(tmp_path)
|
||||
_seed_entry(root, "old", age_days=90, size=3 * 1024 * 1024)
|
||||
|
||||
monkeypatch.setattr(ret, "load_config", lambda: cfg)
|
||||
monkeypatch.setattr(ret.shutil, "disk_usage", lambda _p: _fake_usage(free_gb=200))
|
||||
|
||||
result = CliRunner().invoke(cli, ["retention", "run"])
|
||||
assert result.exit_code == 0, result.output
|
||||
assert "Purged 1 cargo cache entry (3 MB)" in result.output
|
||||
|
||||
|
||||
def test_retention_what_if_keeps_cargo_cache_entries(
|
||||
tmp_path: Path, monkeypatch: pytest.MonkeyPatch
|
||||
) -> None:
|
||||
root, cfg = _cache_config(tmp_path)
|
||||
old = _seed_entry(root, "old", age_days=90)
|
||||
|
||||
monkeypatch.setattr(ret, "load_config", lambda: cfg)
|
||||
monkeypatch.setattr(ret.shutil, "disk_usage", lambda _p: _fake_usage(free_gb=200))
|
||||
|
||||
result = CliRunner().invoke(cli, ["retention", "run", "--what-if"])
|
||||
assert result.exit_code == 0, result.output
|
||||
assert old.exists()
|
||||
assert "would purge 1 cargo cache entry" in result.output
|
||||
|
||||
|
||||
def test_retention_skips_the_cache_when_disabled(
|
||||
tmp_path: Path, monkeypatch: pytest.MonkeyPatch
|
||||
) -> None:
|
||||
root, cfg = _cache_config(tmp_path, enabled=False)
|
||||
old = _seed_entry(root, "old", age_days=90)
|
||||
|
||||
monkeypatch.setattr(ret, "load_config", lambda: cfg)
|
||||
monkeypatch.setattr(ret.shutil, "disk_usage", lambda _p: _fake_usage(free_gb=200))
|
||||
|
||||
result = CliRunner().invoke(cli, ["retention", "run"])
|
||||
assert result.exit_code == 0, result.output
|
||||
assert old.exists(), "a disabled cache must not be touched"
|
||||
assert "disabled" in result.output
|
||||
|
||||
|
||||
def test_low_disk_tightens_the_cache_window_too(
|
||||
tmp_path: Path, monkeypatch: pytest.MonkeyPatch
|
||||
) -> None:
|
||||
"""Aggressive mode must reach the cache — it is the heaviest thing here."""
|
||||
root, cfg = _cache_config(tmp_path, days=30)
|
||||
entry = _seed_entry(root, "old", age_days=10) # young for 30d, old for 7d
|
||||
|
||||
monkeypatch.setattr(ret, "load_config", lambda: cfg)
|
||||
monkeypatch.setattr(ret.shutil, "disk_usage", lambda _p: _fake_usage(free_gb=10))
|
||||
|
||||
result = CliRunner().invoke(cli, ["retention", "run"])
|
||||
assert result.exit_code == 0, result.output
|
||||
assert not entry.exists(), "aggressive retention should have reached the cache"
|
||||
|
||||
|
||||
def test_retention_missing_cache_store_is_not_an_error(
|
||||
tmp_path: Path, monkeypatch: pytest.MonkeyPatch
|
||||
) -> None:
|
||||
_root, cfg = _cache_config(tmp_path)
|
||||
monkeypatch.setattr(ret, "load_config", lambda: cfg)
|
||||
monkeypatch.setattr(ret.shutil, "disk_usage", lambda _p: _fake_usage(free_gb=200))
|
||||
|
||||
result = CliRunner().invoke(cli, ["retention", "run"])
|
||||
assert result.exit_code == 0, result.output
|
||||
assert "nothing to purge" in result.output
|
||||
|
||||
Reference in New Issue
Block a user