navidrome/persistence/artwork_queue_repository.go
Deluan Quintão 07b6411c0b
perf(artwork): cap the stale-absent recheck at 100 items per kind per hour (#6007)
* feat(artwork): drip the stale-absent recheck instead of bursting it daily

Each hourly housekeeping tick now re-queues at most 100 absent states
per kind, oldest attempts first, instead of everything older than 24h
at once. External agents see a flat ~100 requests/hour per agent
instead of hourly bursts of ~2,000, and the effective recheck interval
self-scales with the size of the absent pool (~4 days at 10k absent
artists) while small libraries keep the 24h floor.

* feat(artwork): trust an absent artwork state for a week before rechecking

With the recheck now dripped at 100 items per kind per hour, the 24h
floor only governed small libraries, where the drip cap never binds;
they still re-asked every agent daily. A 7-day floor cuts that cost 7x
and, for large libraries, becomes the binding limit over the drip
cycle (~5.7k calls/day instead of ~9.6k at 10k absent artists).

Among comparable servers, this is still the second-most-eager recheck:
gonic retries misses every 30 days, Jellyfin and Funkwhale never do.

* refactor(artwork): state the drip's backpressure contract where it bites

Review follow-ups: the recheck limit deliberately caps the *selection*,
not the insertions — already-queued rows use up budget, so a stalled
drain admits no new work instead of building a recovery burst. Say so
in the interface doc, mirror it in the mock by truncating the sorted
candidates (matching the SQL's LIMIT-before-ON CONFLICT), and teach
`artwork status` and the worker doc the post-drip wording. Also pin
the one cmd fixture that still assumed a 24h recheck window.
2026-08-21 15:23:07 -04:00

244 lines
9.3 KiB
Go

package persistence
import (
"cmp"
"context"
"fmt"
"slices"
"strings"
"time"
. "github.com/Masterminds/squirrel"
"github.com/navidrome/navidrome/model"
"github.com/navidrome/navidrome/utils/slice"
"github.com/pocketbase/dbx"
)
// Keeps each multi-row insert under SQLite's bind-variable limit (at most 7 vars per row).
const enqueueChunkSize = 100
// Every insert writes these, in this order; the INSERT..SELECT forms must project them to match.
// DequeueBatch also selects exactly these, to leave the drain's rows free of the trace it never reads.
var enqueueColumns = []string{"item_kind", "item_id", "image_type", "priority", "attempts", "retry_at", "enqueued_at"}
type artworkQueueRepository struct {
sqlRepository
}
func NewArtworkQueueRepository(ctx context.Context, db dbx.Builder) model.ArtworkQueueRepository {
r := &artworkQueueRepository{}
r.ctx = ctx
r.db = db
r.tableName = "artwork_queue"
return r
}
func (r *artworkQueueRepository) Get(kind model.Kind, id, imageType string) (*model.ArtworkQueueItem, error) {
var res model.ArtworkQueueItem
err := r.queryOne(Select("*").From(r.tableName).
Where(Eq{"item_kind": kind.Prefix(), "item_id": id, "image_type": imageType}), &res)
if err != nil {
return nil, err
}
return &res, nil
}
// Enqueue starts a fresh lifecycle: it resets enqueued_at (so a fresh request does not inherit an old
// row's spent retry budget) and clears trace (so explain does not show a prior failure at attempts 0).
func (r *artworkQueueRepository) Enqueue(items ...model.ArtworkQueueItem) error {
return r.enqueue(`ON CONFLICT (item_kind, item_id, image_type) DO UPDATE SET
priority = MAX(priority, excluded.priority), retry_at = excluded.retry_at,
attempts = 0, enqueued_at = excluded.enqueued_at, trace = '[]'`, items)
}
func (r *artworkQueueRepository) EnqueuePreservingBackoff(items ...model.ArtworkQueueItem) error {
return r.enqueue(`ON CONFLICT (item_kind, item_id, image_type) DO UPDATE SET
priority = MAX(priority, excluded.priority)`, items)
}
func (r *artworkQueueRepository) EnqueueStaleAbsent(kind model.Kind, attemptedBefore time.Time, limit int) (int64, error) {
now := time.Now()
return r.insertIfNotQueued("", `SELECT item_kind, item_id, image_type, ?, 0, ?, ?
FROM `+itemArtworkTable+` WHERE item_kind = ? AND hash = '' AND attempted_at < ?
ORDER BY attempted_at LIMIT ?`,
model.ArtworkPriorityRecheck, now, now, kind.Prefix(), attemptedBefore, limit)
}
func (r *artworkQueueRepository) EnqueueAllMissing(kind model.Kind, priority int) (int64, error) {
entityTable, ok := artworkOwnerTables[kind]
if !ok {
return 0, fmt.Errorf("artwork queue: no entity table for kind %q", kind.Prefix())
}
now := time.Now()
return r.insertIfNotQueued("", `SELECT ?, id, ?, ?, 0, ?, ?
FROM `+entityTable+`
WHERE id NOT IN (SELECT item_id FROM `+itemArtworkTable+` WHERE item_kind = ?)`,
kind.Prefix(), model.ImageTypePrimary, priority, now, now, kind.Prefix())
}
func (r *artworkQueueRepository) EnqueueIfMissing(items ...model.ArtworkQueueItem) error {
now := time.Now()
for chunk := range slices.Chunk(items, enqueueChunkSize) {
rows := make([]string, 0, len(chunk))
args := make([]any, 0, len(chunk)*4+2)
for _, it := range chunk {
rows = append(rows, "(?,?,?,?)")
args = append(args, it.ItemKind, it.ItemID, cmp.Or(it.ImageType, model.ImageTypePrimary), it.Priority)
}
args = append(args, now, now)
_, err := r.insertIfNotQueued(
`WITH new_items(item_kind, item_id, image_type, priority) AS (VALUES `+strings.Join(rows, ",")+`) `,
`SELECT n.item_kind, n.item_id, n.image_type, n.priority, 0, ?, ?
FROM new_items n
WHERE NOT EXISTS (
SELECT 1 FROM `+itemArtworkTable+` ia
WHERE ia.item_kind = n.item_kind AND ia.item_id = n.item_id AND ia.image_type = n.image_type)`,
args...)
if err != nil {
return err
}
}
return nil
}
// DO NOTHING is deliberate: a recheck must not bump the priority or retry_at of an already-queued item.
const skipIfQueued = ` ON CONFLICT (item_kind, item_id, image_type) DO NOTHING`
// insertIfNotQueued inserts the rows selected by the given SQL, optionally prefixed by a CTE.
func (r *artworkQueueRepository) insertIfNotQueued(with, sql string, args ...any) (int64, error) {
return r.executeSQL(Expr(with+`INSERT INTO `+r.tableName+
` (`+strings.Join(enqueueColumns, ", ")+`) `+sql+skipIfQueued, args...))
}
// artworkSourceFilter selects item_artwork rows of a kind; no sources means every source, "" the absent state.
func artworkSourceFilter(kind model.Kind, sources []string) Sqlizer {
f := And{Eq{"item_kind": kind.Prefix()}}
if len(sources) > 0 {
f = append(f, Eq{"source": sources})
}
return f
}
func (r *artworkQueueRepository) CountBySource(kind model.Kind, sources []string) (int64, error) {
var res struct{ Count int64 }
err := r.queryOne(Select("count(*) as count").From(itemArtworkTable).
Where(artworkSourceFilter(kind, sources)), &res)
return res.Count, err
}
func (r *artworkQueueRepository) SourcesInUse(kind model.Kind) ([]string, error) {
var res []struct{ Source string }
err := r.queryAll(Select("distinct source").From(itemArtworkTable).
Where(Eq{"item_kind": kind.Prefix()}), &res)
if err != nil {
return nil, err
}
return slice.Map(res, func(s struct{ Source string }) string { return s.Source }), nil
}
// EnqueueBySource deliberately leaves item_artwork alone: clearing state in bulk would blank the
// library's artwork until every item is resolved again.
func (r *artworkQueueRepository) EnqueueBySource(kind model.Kind, sources []string, priority int) (int64, error) {
now := time.Now()
sel := Select("item_kind", "item_id", "image_type").
Column(Expr("?", priority)).Column("0").Column(Expr("?", now)).Column(Expr("?", now)).
From(itemArtworkTable).Where(artworkSourceFilter(kind, sources))
return r.executeSQL(Insert(r.tableName).Columns(enqueueColumns...).Select(sel).Suffix(skipIfQueued))
}
func (r *artworkQueueRepository) enqueue(conflict string, items []model.ArtworkQueueItem) error {
now := time.Now()
for chunk := range slices.Chunk(items, enqueueChunkSize) {
ins := Insert(r.tableName).Columns(enqueueColumns...)
for _, it := range chunk {
ins = ins.Values(it.ItemKind, it.ItemID, cmp.Or(it.ImageType, model.ImageTypePrimary), it.Priority, 0, now, now)
}
ins = ins.Suffix(conflict)
if _, err := r.executeSQL(ins); err != nil {
return err
}
}
return nil
}
func (r *artworkQueueRepository) DequeueBatch(n int, kinds ...string) ([]model.ArtworkQueueItem, error) {
sel := Select(enqueueColumns...).From(r.tableName).
Where(LtOrEq{"retry_at": time.Now()}).
OrderBy("priority DESC", "enqueued_at ASC").
Limit(uint64(n))
if len(kinds) > 0 {
sel = sel.Where(Eq{"item_kind": kinds})
}
var res []model.ArtworkQueueItem
err := r.queryAll(sel, &res)
return res, err
}
func (r *artworkQueueRepository) MarkFailedIfUnchanged(kind, id, imageType string, seenRetryAt, retryAt time.Time, trace string) error {
upd := Update(r.tableName).
Set("attempts", Expr("attempts + 1")).
Set("retry_at", retryAt).
Set("trace", trace).
Where(Eq{"item_kind": kind, "item_id": id, "image_type": imageType, "retry_at": seenRetryAt})
_, err := r.executeSQL(upd)
return err
}
func (r *artworkQueueRepository) DeleteIfUnchanged(kind, id, imageType string, retryAt time.Time) error {
return r.delete(Eq{"item_kind": kind, "item_id": id, "image_type": imageType, "retry_at": retryAt})
}
func (r *artworkQueueRepository) PurgeDangling() (int64, error) {
return purgeDangling(r.sqlRepository)
}
// artworkQueueFilter returns no conditions for an empty filter, so an unfiltered DELETE keeps
// SQLite's truncate path. It ignores retry_at: a backing-off row is pending work too.
func artworkQueueFilter(kinds []model.Kind, priorities []int) And {
var f And
if len(kinds) > 0 {
f = append(f, Eq{"item_kind": model.KindPrefixes(kinds)})
}
if len(priorities) > 0 {
f = append(f, Eq{"priority": priorities})
}
return f
}
// CountQueued shares its filter with PurgeQueued, so a preview cannot count rows the delete misses.
func (r *artworkQueueRepository) CountQueued(kinds []model.Kind, priorities []int) ([]model.ArtworkQueueStat, error) {
sel := Select("item_kind", "priority", "count(*) as count").From(r.tableName).
GroupBy("item_kind", "priority").OrderBy("item_kind", "priority desc")
if f := artworkQueueFilter(kinds, priorities); len(f) > 0 {
sel = sel.Where(f)
}
var res []model.ArtworkQueueStat
err := r.queryAll(sel, &res)
return res, err
}
func (r *artworkQueueRepository) PurgeQueued(kinds []model.Kind, priorities []int) (int64, error) {
del := Delete(r.tableName)
if f := artworkQueueFilter(kinds, priorities); len(f) > 0 {
del = del.Where(f)
}
return r.executeSQL(del)
}
func (r *artworkQueueRepository) Count() (int64, error) {
var res struct{ Count int64 }
err := r.queryOne(Select("count(*) as count").From(r.tableName), &res)
return res.Count, err
}
// CountAbsent matches EnqueueStaleAbsent on hash, so the stale count is the pool a recheck drains from.
func (r *artworkQueueRepository) CountAbsent(kind model.Kind, attemptedBefore time.Time) (model.ArtworkAbsentStat, error) {
var res model.ArtworkAbsentStat
err := r.queryOne(Select("count(*) as total").
Column(Expr("coalesce(sum(attempted_at < ?), 0) as stale", attemptedBefore)).
From(itemArtworkTable).Where(Eq{"item_kind": kind.Prefix(), "hash": ""}), &res)
return res, err
}
var _ model.ArtworkQueueRepository = (*artworkQueueRepository)(nil)