Merge pull request 'fix(retry): глобальный rate-limit + Retry-After + cooldown (#70)' (#71) from fix/oo-backoff#70 into main

AGENTS.md

@@ -9,7 +9,7 @@ Canonical Go client for OnlyOffice Workspace (Projects + Calendar + CRM) and the
   - `request.go` — `Request`, `Query`, `Time`, `Token`, `MetaResponse`, `Permissions`.
   - `auth.go` — `Authenticate`, `AuthenticateContext`, `InvalidateToken`, `Auth`, token lifecycle.
   - `http.go` — transport + DRY response decoders (`ResponseArray`/`ResponseObject`/`postFormObject`/`putFormObject`/`deleteObject`).
-  - `projects.go`, `tasks.go`, `users.go`, `calendar.go`, `crm.go`, `files.go`, `files_webdav.go`, `files_stem.go`, `retry.go`, `mails.go`, `invoices.go` — typed / untyped domain methods. **`files.go`** — CRM opportunity upload plus **project/task Documents** (`UpdateFile`, `UploadToFolderReplacing`). **`files_webdav.go`** — Documents module by id (`ListDavFolder`, `MoveDavItems`/`CopyDavItems` with per-operation error surfacing, `ListFileOps`). **`retry.go`** — `DoRetry`: deterministic linear backoff (no jitter) on 429/502/503/504; every bulk tool routes API calls through it, and the HTTP transport + auth (`retryRaw`, `AuthenticateContext`) retry transient answers centrally. **`mails.go`** — OnlyOffice Workspace Mail. **`invoices.go`** — CRM invoices, PDF regen/cleanup, status. Association rules: [`docs/crm-associations.md`](docs/crm-associations.md).
+  - `projects.go`, `tasks.go`, `users.go`, `calendar.go`, `crm.go`, `files.go`, `files_webdav.go`, `files_stem.go`, `retry.go`, `mails.go`, `invoices.go` — typed / untyped domain methods. **`files.go`** — CRM opportunity upload plus **project/task Documents** (`UpdateFile`, `UploadToFolderReplacing`). **`files_webdav.go`** — Documents module by id (`ListDavFolder`, `MoveDavItems`/`CopyDavItems` with per-operation error surfacing, `ListFileOps`). **`retry.go`** — `DoRetry`: deterministic exponential backoff (no jitter) on 429/502/503/504; every bulk tool routes API calls through it, and the HTTP transport + auth (`retryRaw`, `AuthenticateContext`) retry transient answers centrally. `ratelimit.go` adds a process-wide token bucket (`OO_RATE_LIMIT`/`OO_BURST`) and a shared 429 cooldown gate, installed via `pacedTransport` in `NewClient`; `Retry-After` is parsed into `*TransientError` and honoured. See [`docs/rate-limiting.md`](docs/rate-limiting.md). **`mails.go`** — OnlyOffice Workspace Mail. **`invoices.go`** — CRM invoices, PDF regen/cleanup, status. Association rules: [`docs/crm-associations.md`](docs/crm-associations.md).
   - **Unified file client (epic #34) — `file_core.go`, `file_rest.go`, `file_dav.go`, `file_pg.go`, `file_es.go`, `file_es_text.go`, `file_text_index.go`, `file_facade.go`.** `file_core.go` — model (`Entry`, `Kind`) + `FileStore`/`Searcher`; `file_rest.go`/`file_dav.go` — REST/WebDAV adapters; `file_pg.go` — **read-only** SQL store (PostgreSQL/MySQL, `ErrReadOnly` on writes); `file_es.go` — OnlyOffice Elasticsearch searcher; `file_es_text.go`/`file_text_index.go` — own PDF/scan index (`oo_docs_text`, PDF attachments via pdfdetach); `file_facade.go` — `FileClient` with read/write/search order and transient fallback. Use `c.Files()` (facade), `c.FileStore("rest"|"dav"|"pg"|"sql")` or `c.SQLFileStore()`; contract and how to add a backend: [`docs/unified-file-client.md`](docs/unified-file-client.md).
   - Pure stdlib + `google/go-querystring`; no UI, no dotenv.
 - **CLI — `cmd/oo/` as `package main`.** Cobra wrapper that loads `.env` via `godotenv` at startup. **Subject-based command tree** mirroring [`tea`](https://gitea.com/gitea/tea):

CHANGELOG.md

@@ -6,6 +6,14 @@ adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ## Unreleased
 
+
+### Features
+
+* **retry:** global token-bucket rate limit (`OO_RATE_LIMIT`/`OO_BURST`), typed
+  `TransientError` with `Retry-After`, exponential backoff
+  (`OO_RETRY_ATTEMPTS`/`_BASE`/`_MAX`) and a process-wide 429 cooldown gate.
+  All HTTP paths are paced via `pacedTransport`.
+
 ## [0.18.0](https://github.com/eSlider/go-onlyoffice/compare/v0.17.0...v0.18.0) (2026-09-04)
 
 

README.md

@@ -522,9 +522,9 @@ type Task struct {
 | `ListFileOps(ctx)` | Active file operations (move/copy status polling) |
 | `FolderFiles(ctx, folderID)` | Flat file list of a folder (stem helpers) |
 | `DeleteFilesByStem(ctx, folderID, stem)` | Remove `stem\|ext` copies |
-| `DoRetry(ctx, policy, fn)` | Deterministic linear backoff (N·Base, no jitter) on 429/502/503/504 |
-| `DefaultRetryPolicy()` | 5 attempts, 1s·2s·3s·4s waits, 30s cap |
-| `Transient(err)` | True for retriable OnlyOffice answers |
+| `DoRetry(ctx, policy, fn)` | Deterministic exponential backoff (`Base·2^(N-1)`, no jitter) on 429/502/503/504; honours `Retry-After` and the process-wide cooldown gate |
+| `DefaultRetryPolicy()` | From env: 7 attempts, 2s base, 2m cap (`OO_RETRY_ATTEMPTS/_BASE/_MAX`) |
+| `Transient(err)` | True for retriable OnlyOffice answers (`*TransientError` or HTTP 429/502/503/504 text) |
 
 ### Helper Types
 
@@ -732,9 +732,12 @@ fallback rules, env names and how to add a backend:
 ### Bulk tools (`cmd/`)
 
 Small single-purpose binaries for bulk Documents work. All of them pace
-requests and retry transient OnlyOffice answers (429/502/503/504) with a
-deterministic linear backoff — no jitter, same waits on every run
-(see `DoRetry` below). Build with `go build ./cmd/<tool>`.
+requests through a process-wide token bucket (default ~4 req/s, `OO_RATE_LIMIT`/
+`OO_BURST`) and retry transient OnlyOffice answers (429/502/503/504) with a
+deterministic exponential backoff — no jitter, same waits on every run. A 429
+opens a shared cooldown gate and `Retry-After` is honoured (see `DoRetry` and
+[`docs/rate-limiting.md`](docs/rate-limiting.md)). Build with
+`go build ./cmd/<tool>`.
 
 ```bash
 ooscan 659                             # recursive index → TSV: file_id, folder_id, path, title
@@ -999,6 +1002,11 @@ oo projects files list 33
 | `ONLYOFFICE_CALENDAR_ID` | Default calendar id used when omitted (default `1`) |
 | `ONLYOFFICE_PROJECT_ID` | Default project id used when omitted (default `33`) |
 | `OO_URL`, `OO_USER`, `OO_PASS` | Optional CLI-only aliases for `ONLYOFFICE_*` |
+| `OO_RATE_LIMIT` | Process-wide request pacing, req/s (default `4`; `0` disables) |
+| `OO_BURST` | Token-bucket burst (default `1`) |
+| `OO_RETRY_ATTEMPTS` | Transient retries, total attempts (default `7`) |
+| `OO_RETRY_BASE` | Exponential backoff base (default `2s`) |
+| `OO_RETRY_MAX` | Backoff cap (default `2m`) |
 | `ONLYOFFICE_ES_URL` | Elasticsearch base URL (`oo search`, own index); see [`docs/elasticsearch.md`](docs/elasticsearch.md) |
 | `ONLYOFFICE_ES_INDEX` | OnlyOffice index (default `files_file`) |
 | `ONLYOFFICE_ES_TEXT_INDEX` | Own PDF/scan index (default `oo_docs_text`) |

auth.go

@@ -83,7 +83,7 @@ func (c *Client) authenticateOnce(ctx context.Context) error {
 		return err
 	}
 	if resp.StatusCode >= 400 {
-		return fmt.Errorf("auth: %d %s", resp.StatusCode, truncate(string(raw), 400))
+		return statusError(resp.StatusCode, retryAfterOf(resp), "auth: %d %s", resp.StatusCode, truncate(string(raw), 400))
 	}
 	var env struct {
 		Response *Token `json:"response"`

client.go

@@ -38,11 +38,15 @@ type Client struct {
 	folderTitlesMu sync.Mutex
 }
 
-// NewClient returns a new Client backed by http.DefaultClient.
+// NewClient returns a new Client whose transport is paced by the process-wide
+// rate limiter and 429 cooldown gate (OO_RATE_LIMIT/OO_BURST; see ratelimit.go).
 func NewClient(c Credentials) *Client {
 	jar, _ := cookiejar.New(nil)
 	return &Client{
-		client:      &http.Client{Jar: jar},
+		client: &http.Client{
+			Jar:       jar,
+			Transport: &pacedTransport{base: http.DefaultTransport},
+		},
 		credentials: &c,
 	}
 }

docs/README.md

@@ -22,6 +22,9 @@ related:
 - [rclone-webdav.md](rclone-webdav.md) — rclone-монтирование Documents
   (`deploy/docker-compose.rclone-webdav.yml`), smoke, ограничения.
 - [crm-associations.md](crm-associations.md) — правила ассоциаций CRM.
+- [rate-limiting.md](rate-limiting.md) — rate limit, exponential backoff,
+  `Retry-After`, общий cooldown против 429; env `OO_RATE_LIMIT`/`OO_BURST`/
+  `OO_RETRY_*`.
 
 ## Тесты
 Команды и туннели — раздел Testing в [README.md](../README.md#testing).

docs/rate-limiting.md

@@ -0,0 +1,55 @@
+---
+type: reference
+status: current
+related:
+  - README.md
+  - ../AGENTS.md
+---
+
+# Rate limit, backoff и cooldown
+
+Устойчивость к 429 (openresty). Всё встроено в библиотеку — отдельный пакет не
+нужен. Реализация: `ratelimit.go`, `retry.go`.
+
+## Что происходит с каждым запросом
+
+1. **Cooldown-гейт** — общий на процесс. Если недавно пришёл 429, все запросы
+   ждут конца окна.
+2. **Rate limiter** — token bucket на процесс. Пейсит все HTTP-пути: листинг,
+   создание папок, загрузку, `get project`, auth.
+3. Запрос уходит.
+4. Ответ ≥400 → `*TransientError` (для 429/502/503/504) с `Retry-After`.
+5. `DoRetry` — экспоненциальный backoff, без jitter.
+6. `Retry-After` длиннее backoff → ждём его; окно уходит в общий cooldown.
+
+Установлено в `NewClient` через `pacedTransport`; отдельный код трогать не надо.
+
+## Env
+
+| Переменная | Default | Смысл |
+|---|---|---|
+| `OO_RATE_LIMIT` | `4` | запросов/с на процесс; `0` — лимитер выключен |
+| `OO_BURST` | `1` | запас токенов token bucket |
+| `OO_RETRY_ATTEMPTS` | `7` | всего попыток, включая первую |
+| `OO_RETRY_BASE` | `2s` | база экспоненты: ждать перед попыткой N = `Base*2^(N-1)` |
+| `OO_RETRY_MAX` | `2m` | потолок ожидания |
+
+Битые значения → default. `OO_RETRY_*` — формат `time.ParseDuration`
+(`2s`, `30s`, `2m`).
+
+## Правила
+
+- Детерминированно, без jitter — повторный прогон ждёт столько же.
+- `Retry-After` — секунды (`120`) или HTTP-date.
+- Cooldown общий: параллельные и последовательные вызовы не бьют в стену.
+- Backoff cap не ограничивает `Retry-After` — серверу верим больше.
+- Только stdlib.
+
+## Когда руками снять нагрузку
+
+`OO_RATE_LIMIT` ниже (`2`), `OO_BURST=1`; при массовом apply — батчами.
+
+## Тесты
+
+`retry_test.go` — `Retry-After`, экспонента, cap; `ratelimit_test.go` — burst,
+`OO_RATE_LIMIT=0`, cooldown. Фейковый сервер отдаёт 429 с заголовком.

files_webdav.go

@@ -325,7 +325,7 @@ func (c *Client) deleteJSON(ctx context.Context, path string, body any) (json.Ra
 		return nil, err
 	}
 	if resp.StatusCode >= 400 {
-		return nil, fmt.Errorf("DELETE %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
+		return nil, statusError(resp.StatusCode, retryAfterOf(resp), "DELETE %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
 	}
 	return raw, nil
 }
@@ -365,7 +365,7 @@ func (c *Client) uploadReader(ctx context.Context, path, fieldName, fileName str
 		return nil, err
 	}
 	if resp.StatusCode >= 400 {
-		return nil, fmt.Errorf("upload %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
+		return nil, statusError(resp.StatusCode, retryAfterOf(resp), "upload %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
 	}
 	return raw, nil
 }

http.go

@@ -171,7 +171,7 @@ func (c *Client) getJSONOnce(ctx context.Context, path string) (json.RawMessage,
 		return nil, err
 	}
 	if resp.StatusCode >= 400 {
-		return nil, fmt.Errorf("GET %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
+		return nil, statusError(resp.StatusCode, retryAfterOf(resp), "GET %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
 	}
 	return raw, nil
 }
@@ -219,7 +219,7 @@ func (c *Client) formRequestOnce(ctx context.Context, method, path string, field
 		return nil, err
 	}
 	if resp.StatusCode >= 400 {
-		return nil, fmt.Errorf("%s form %s: %d %s", method, path, resp.StatusCode, truncate(string(raw), 400))
+		return nil, statusError(resp.StatusCode, retryAfterOf(resp), "%s form %s: %d %s", method, path, resp.StatusCode, truncate(string(raw), 400))
 	}
 	return raw, nil
 }
@@ -250,7 +250,7 @@ func (c *Client) deleteReqOnce(ctx context.Context, path string) (json.RawMessag
 		return nil, err
 	}
 	if resp.StatusCode >= 400 {
-		return nil, fmt.Errorf("DELETE %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
+		return nil, statusError(resp.StatusCode, retryAfterOf(resp), "DELETE %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
 	}
 	return raw, nil
 }
@@ -315,7 +315,7 @@ func (c *Client) postJSONOnce(ctx context.Context, path string, body any) (json.
 		return nil, err
 	}
 	if resp.StatusCode >= 400 {
-		return nil, fmt.Errorf("POST JSON %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
+		return nil, statusError(resp.StatusCode, retryAfterOf(resp), "POST JSON %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
 	}
 	return raw, nil
 }
@@ -362,7 +362,7 @@ func (c *Client) putJSONOnce(ctx context.Context, path string, body any) (json.R
 		return nil, err
 	}
 	if resp.StatusCode >= 400 {
-		return nil, fmt.Errorf("PUT JSON %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
+		return nil, statusError(resp.StatusCode, retryAfterOf(resp), "PUT JSON %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
 	}
 	return raw, nil
 }
@@ -423,7 +423,7 @@ func (c *Client) uploadMultipartOnce(ctx context.Context, method, path, fieldNam
 		return nil, err
 	}
 	if resp.StatusCode >= 400 {
-		return nil, fmt.Errorf("upload %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
+		return nil, statusError(resp.StatusCode, retryAfterOf(resp), "upload %s: %d %s", path, resp.StatusCode, truncate(string(raw), 400))
 	}
 	return raw, nil
 }

mails.go

@@ -125,7 +125,7 @@ func (c *Client) DownloadMailAttachment(ctx context.Context, attachmentID string
 		return nil, err
 	}
 	if resp.StatusCode >= 400 {
-		return nil, fmt.Errorf("DownloadMailAttachment %s: %d %s", id, resp.StatusCode, truncate(string(raw), 400))
+		return nil, statusError(resp.StatusCode, retryAfterOf(resp), "DownloadMailAttachment %s: %d %s", id, resp.StatusCode, truncate(string(raw), 400))
 	}
 	return raw, nil
 }

ratelimit.go

@@ -0,0 +1,210 @@
+package onlyoffice
+
+// Client-side pacing and the process-wide 429 gate. A token bucket applies to
+// every HTTP request a *Client makes (via pacedTransport), so listing, folder
+// creation, uploads and even `get project` share one budget and cannot burst
+// past openresty. A typed transient answer carrying Retry-After opens the
+// cooldown gate for the whole process.
+
+import (
+	"context"
+	"net/http"
+	"os"
+	"strconv"
+	"strings"
+	"sync"
+	"time"
+)
+
+// Default pacing: ~4 requests/second with a single-request burst. This keeps
+// bulk syncs under openresty's threshold without noticeably slowing one-shot
+// tools.
+const (
+	defaultRateLimit = 4.0
+	defaultBurst     = 1
+)
+
+// rateLimiter is a deterministic token bucket. A nil *rateLimiter is a no-op,
+// which is how OO_RATE_LIMIT=0 disables pacing.
+type rateLimiter struct {
+	mu     sync.Mutex
+	rate   float64
+	burst  float64
+	tokens float64
+	last   time.Time
+}
+
+// newRateLimiter returns a limiter at rate requests/second with the given
+// burst. rate <= 0 disables pacing and returns nil.
+func newRateLimiter(rate float64, burst int) *rateLimiter {
+	if rate <= 0 {
+		return nil
+	}
+	if burst < 1 {
+		burst = 1
+	}
+	return &rateLimiter{
+		rate:   rate,
+		burst:  float64(burst),
+		tokens: float64(burst),
+		last:   time.Now(),
+	}
+}
+
+// wait consumes one token, blocking until one is available or ctx is done.
+func (l *rateLimiter) wait(ctx context.Context) error {
+	if l == nil {
+		return nil
+	}
+	for {
+		l.mu.Lock()
+		now := time.Now()
+		l.tokens += now.Sub(l.last).Seconds() * l.rate
+		if l.tokens > l.burst {
+			l.tokens = l.burst
+		}
+		l.last = now
+		if l.tokens >= 1 {
+			l.tokens--
+			l.mu.Unlock()
+			return nil
+		}
+		need := time.Duration((1 - l.tokens) / l.rate * float64(time.Second))
+		l.mu.Unlock()
+		if need < time.Millisecond {
+			need = time.Millisecond
+		}
+		select {
+		case <-ctx.Done():
+			return ctx.Err()
+		case <-time.After(need):
+		}
+	}
+}
+
+// cooldownGate is the process-wide 429 gate: after a throttled answer every
+// request waits until the gate opens, so concurrent and sequential callers do
+// not hammer the server.
+type cooldownGate struct {
+	mu    sync.Mutex
+	until time.Time
+}
+
+// note extends the gate to at least now+d.
+func (g *cooldownGate) note(d time.Duration) {
+	if d <= 0 {
+		return
+	}
+	g.mu.Lock()
+	if u := time.Now().Add(d); u.After(g.until) {
+		g.until = u
+	}
+	g.mu.Unlock()
+}
+
+// wait blocks until the gate opens or ctx is done.
+func (g *cooldownGate) wait(ctx context.Context) error {
+	g.mu.Lock()
+	d := time.Until(g.until)
+	g.mu.Unlock()
+	if d <= 0 {
+		return nil
+	}
+	t := time.NewTimer(d)
+	defer t.Stop()
+	select {
+	case <-ctx.Done():
+		return ctx.Err()
+	case <-t.C:
+		return nil
+	}
+}
+
+var (
+	globalLimiter   *rateLimiter
+	globalLimiterOn bool
+	globalLimiterMu sync.Mutex
+	globalCooldown  = &cooldownGate{}
+)
+
+// limiter returns the process-wide rate limiter, initialising it from the
+// environment on first use. OO_RATE_LIMIT=0 yields a nil (disabled) limiter.
+func limiter() *rateLimiter {
+	globalLimiterMu.Lock()
+	defer globalLimiterMu.Unlock()
+	if !globalLimiterOn {
+		globalLimiter = newRateLimiterFromEnv()
+		globalLimiterOn = true
+	}
+	return globalLimiter
+}
+
+// newRateLimiterFromEnv builds a limiter from OO_RATE_LIMIT (requests/second,
+// default 4; 0 disables) and OO_BURST (default 1). Malformed values fall back
+// to the defaults.
+func newRateLimiterFromEnv() *rateLimiter {
+	rate := defaultRateLimit
+	if v, ok := os.LookupEnv("OO_RATE_LIMIT"); ok {
+		if f, err := strconv.ParseFloat(strings.TrimSpace(v), 64); err == nil {
+			rate = f
+		}
+	}
+	burst := defaultBurst
+	if v, ok := os.LookupEnv("OO_BURST"); ok {
+		if n, err := strconv.Atoi(strings.TrimSpace(v)); err == nil {
+			burst = n
+		}
+	}
+	return newRateLimiter(rate, burst)
+}
+
+// pacedTransport applies the process-wide cooldown gate and rate limiter to
+// every request, and arms the gate on a 429. Installed by NewClient, so all
+// HTTP paths — typed Query, the http.go helpers, auth and WebDAV — are paced.
+type pacedTransport struct {
+	base http.RoundTripper
+}
+
+func (t *pacedTransport) RoundTrip(req *http.Request) (*http.Response, error) {
+	if err := globalCooldown.wait(req.Context()); err != nil {
+		return nil, err
+	}
+	if err := limiter().wait(req.Context()); err != nil {
+		return nil, err
+	}
+	resp, err := t.base.RoundTrip(req)
+	if err != nil {
+		return nil, err
+	}
+	if resp.StatusCode == http.StatusTooManyRequests {
+		globalCooldown.note(retryAfterOf(resp))
+	}
+	return resp, nil
+}
+
+// envInt reads a positive integer env var, falling back to def.
+func envInt(name string, def int) int {
+	v, ok := os.LookupEnv(name)
+	if !ok {
+		return def
+	}
+	n, err := strconv.Atoi(strings.TrimSpace(v))
+	if err != nil || n <= 0 {
+		return def
+	}
+	return n
+}
+
+// envDuration reads a duration env var (Go syntax, e.g. "2s", "2m"),
+// falling back to def.
+func envDuration(name string, def time.Duration) time.Duration {
+	v, ok := os.LookupEnv(name)
+	if !ok {
+		return def
+	}
+	d, err := time.ParseDuration(strings.TrimSpace(v))
+	if err != nil || d <= 0 {
+		return def
+	}
+	return d
+}

ratelimit_test.go

@@ -0,0 +1,160 @@
+package onlyoffice
+
+import (
+	"context"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"testing"
+	"time"
+)
+
+// TestMain disables the process-wide rate limiter for the suite so existing
+// fake-server tests stay fast; the pacing tests configure their own limiters.
+func TestMain(m *testing.M) {
+	_ = os.Setenv("OO_RATE_LIMIT", "0")
+	os.Exit(m.Run())
+}
+
+// resetPacingForTest restores the package-level limiter and cooldown gate to a
+// clean state. White-box helper for tests that arm the global 429 gate.
+func resetPacingForTest() {
+	globalLimiterMu.Lock()
+	globalLimiter = nil
+	globalLimiterOn = false
+	globalLimiterMu.Unlock()
+	globalCooldown = &cooldownGate{}
+}
+
+func TestNewClientUsesPacedTransport(t *testing.T) {
+	c := NewClient(Credentials{Url: "https://example.test"})
+	if _, ok := c.client.Transport.(*pacedTransport); !ok {
+		t.Fatalf("transport = %T, want *pacedTransport", c.client.Transport)
+	}
+}
+
+func TestNewRateLimiterFromEnv(t *testing.T) {
+	t.Setenv("OO_RATE_LIMIT", "")
+	t.Setenv("OO_BURST", "")
+	if l := newRateLimiterFromEnv(); l == nil || l.rate != defaultRateLimit || l.burst != defaultBurst {
+		t.Fatalf("defaults: got %+v, want rate=%v burst=%d", l, defaultRateLimit, defaultBurst)
+	}
+
+	t.Setenv("OO_RATE_LIMIT", "0")
+	if l := newRateLimiterFromEnv(); l != nil {
+		t.Fatalf("OO_RATE_LIMIT=0: got %+v, want nil (off)", l)
+	}
+
+	t.Setenv("OO_RATE_LIMIT", "10")
+	t.Setenv("OO_BURST", "3")
+	if l := newRateLimiterFromEnv(); l == nil || l.rate != 10 || l.burst != 3 {
+		t.Fatalf("explicit: got %+v, want rate=10 burst=3", l)
+	}
+}
+
+func TestNilRateLimiterIsNoOp(t *testing.T) {
+	var l *rateLimiter
+	start := time.Now()
+	if err := l.wait(context.Background()); err != nil {
+		t.Fatalf("nil limiter wait: %v", err)
+	}
+	if elapsed := time.Since(start); elapsed > 20*time.Millisecond {
+		t.Fatalf("nil limiter blocked for %v", elapsed)
+	}
+}
+
+func TestRateLimiterPacesBurst(t *testing.T) {
+	l := newRateLimiter(50, 1) // 20ms per token, no burst headroom
+	start := time.Now()
+	for i := 0; i < 5; i++ {
+		if err := l.wait(context.Background()); err != nil {
+			t.Fatalf("wait %d: %v", i, err)
+		}
+	}
+	// 5 calls with burst 1 -> 4 gaps * 20ms = 80ms.
+	if elapsed := time.Since(start); elapsed < 60*time.Millisecond {
+		t.Fatalf("elapsed %v, want >= 60ms (not paced)", elapsed)
+	}
+}
+
+func TestRateLimiterHonoursContext(t *testing.T) {
+	l := newRateLimiter(1, 1)
+	if err := l.wait(context.Background()); err != nil {
+		t.Fatalf("first wait: %v", err)
+	}
+	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Millisecond)
+	defer cancel()
+	if err := l.wait(ctx); err == nil {
+		t.Fatal("want context error while waiting for a token")
+	}
+}
+
+func TestCooldownGateBlocksAndExpires(t *testing.T) {
+	g := &cooldownGate{}
+	g.note(50 * time.Millisecond)
+	start := time.Now()
+	if err := g.wait(context.Background()); err != nil {
+		t.Fatalf("wait: %v", err)
+	}
+	if elapsed := time.Since(start); elapsed < 45*time.Millisecond {
+		t.Fatalf("gate did not block: %v", elapsed)
+	}
+	start = time.Now()
+	if err := g.wait(context.Background()); err != nil {
+		t.Fatalf("second wait: %v", err)
+	}
+	if elapsed := time.Since(start); elapsed > 20*time.Millisecond {
+		t.Fatalf("expired gate still blocked: %v", elapsed)
+	}
+}
+
+func TestCooldownGateNoteExtendsNotShortens(t *testing.T) {
+	g := &cooldownGate{}
+	g.note(80 * time.Millisecond)
+	g.note(10 * time.Millisecond) // must not shorten the window
+	if d := time.Until(g.until); d < 70*time.Millisecond {
+		t.Fatalf("note shortened gate: %v left", d)
+	}
+}
+
+// TestPacedTransportArmsCooldownOn429 verifies that a throttled answer opens the
+// process-wide gate and that the next request is blocked for the Retry-After
+// window even though it targets a healthy endpoint.
+func TestPacedTransportArmsCooldownOn429(t *testing.T) {
+	resetPacingForTest()
+	defer resetPacingForTest()
+
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.URL.Path == "/throttle" {
+			w.Header().Set("Retry-After", "1")
+			w.WriteHeader(http.StatusTooManyRequests)
+			return
+		}
+		w.WriteHeader(http.StatusOK)
+		_, _ = w.Write([]byte("{}"))
+	}))
+	defer srv.Close()
+
+	tr := &pacedTransport{base: http.DefaultTransport}
+
+	req1, _ := http.NewRequestWithContext(context.Background(), http.MethodGet, srv.URL+"/throttle", nil)
+	resp1, err := tr.RoundTrip(req1)
+	if err != nil {
+		t.Fatalf("throttled round trip: %v", err)
+	}
+	resp1.Body.Close()
+	if resp1.StatusCode != http.StatusTooManyRequests {
+		t.Fatalf("status = %d, want 429", resp1.StatusCode)
+	}
+
+	start := time.Now()
+	req2, _ := http.NewRequestWithContext(context.Background(), http.MethodGet, srv.URL+"/ok", nil)
+	resp2, err := tr.RoundTrip(req2)
+	if err != nil {
+		t.Fatalf("second round trip: %v", err)
+	}
+	resp2.Body.Close()
+	if elapsed := time.Since(start); elapsed < 900*time.Millisecond {
+		t.Fatalf("cooldown did not block next call: %v", elapsed)
+	}
+}

request.go

@@ -102,7 +102,7 @@ func (c *Client) queryOnce(request Request, result interface{}) error {
 		return err
 	}
 	if resp.StatusCode >= 400 {
-		return fmt.Errorf("%s %s: %d %s", request.GetMethod(), request.Uri, resp.StatusCode, truncate(string(raw), 400))
+		return statusError(resp.StatusCode, retryAfterOf(resp), "%s %s: %d %s", request.GetMethod(), request.Uri, resp.StatusCode, truncate(string(raw), 400))
 	}
 	if result == nil {
 		return nil

retry.go

@@ -3,38 +3,149 @@ package onlyoffice
 import (
 	"context"
 	"encoding/json"
+	"errors"
+	"fmt"
+	"net/http"
 	"regexp"
+	"strconv"
+	"strings"
 	"time"
 )
 
-// RetryPolicy controls deterministic retries against OnlyOffice: fixed linear
+// RetryPolicy controls deterministic retries against OnlyOffice: exponential
 // backoff without jitter, so repeated runs wait exactly the same schedule.
 // OnlyOffice throttles bulk reads/writes with 429 (and occasional 502/503/504
 // from openresty), so every bulk tool routes API calls through DoRetry.
 type RetryPolicy struct {
 	Attempts int           // total attempts, including the first try
-	Base     time.Duration // wait before retry N is N*Base
-	Max      time.Duration // per-wait cap
+	Base     time.Duration // wait before retry N is Base*2^(N-1)
+	Max      time.Duration // per-wait cap; <=0 means no cap
 }
 
-// DefaultRetryPolicy retries up to 5 times with 1s, 2s, 3s, 4s waits.
+// DefaultRetryPolicy builds the policy from the environment, falling back to
+// 7 attempts, a 2s base and a 2m cap:
+//
+//   - OO_RETRY_ATTEMPTS (default 7)
+//   - OO_RETRY_BASE     (duration, default 2s)
+//   - OO_RETRY_MAX      (duration, default 2m)
 func DefaultRetryPolicy() RetryPolicy {
-	return RetryPolicy{Attempts: 5, Base: time.Second, Max: 30 * time.Second}
+	return RetryPolicy{
+		Attempts: envInt("OO_RETRY_ATTEMPTS", 7),
+		Base:     envDuration("OO_RETRY_BASE", 2*time.Second),
+		Max:      envDuration("OO_RETRY_MAX", 2*time.Minute),
+	}
 }
 
 var transientRe = regexp.MustCompile(`:\s*(429|502|503|504)\b`)
 
-// Transient reports whether err looks like a transient OnlyOffice answer
-// (an HTTP 429/502/503/504 surfaced as "...: <code> ...").
+// isTransientStatus reports whether an HTTP status is retriable at the edge.
+func isTransientStatus(status int) bool {
+	switch status {
+	case http.StatusTooManyRequests, http.StatusBadGateway,
+		http.StatusServiceUnavailable, http.StatusGatewayTimeout:
+		return true
+	}
+	return false
+}
+
+// TransientError is a typed transient answer from the HTTP layer. It carries
+// the status code and, when present, the server's Retry-After delay so DoRetry
+// can wait at least that long.
+type TransientError struct {
+	Code       int
+	RetryAfter time.Duration
+	Msg        string
+}
+
+func (e *TransientError) Error() string { return e.Msg }
+
+// Transient reports whether err looks like a transient OnlyOffice answer: a
+// *TransientError with a retriable code, or an error whose text carries an
+// HTTP 429/502/503/504.
 func Transient(err error) bool {
 	if err == nil {
 		return false
 	}
+	var te *TransientError
+	if errors.As(err, &te) {
+		return isTransientStatus(te.Code)
+	}
 	return transientRe.MatchString(err.Error())
 }
 
+// statusError wraps a non-2xx answer, tagging transient statuses so DoRetry
+// recognises them and honours Retry-After.
+func statusError(status int, retryAfter time.Duration, format string, args ...any) error {
+	msg := fmt.Sprintf(format, args...)
+	if isTransientStatus(status) {
+		return &TransientError{Code: status, RetryAfter: retryAfter, Msg: msg}
+	}
+	return errors.New(msg)
+}
+
+// retryAfterOf parses the Retry-After header of a response (integer seconds or
+// an HTTP-date). Returns 0 when absent or malformed.
+func retryAfterOf(resp *http.Response) time.Duration {
+	if resp == nil {
+		return 0
+	}
+	return parseRetryAfter(resp.Header.Get("Retry-After"))
+}
+
+// parseRetryAfter parses a Retry-After value: delay-seconds (RFC 9110) or an
+// HTTP-date. Zero, negative and malformed values yield 0.
+func parseRetryAfter(v string) time.Duration {
+	v = strings.TrimSpace(v)
+	if v == "" {
+		return 0
+	}
+	if secs, err := strconv.Atoi(v); err == nil {
+		if secs <= 0 {
+			return 0
+		}
+		return time.Duration(secs) * time.Second
+	}
+	if t, err := http.ParseTime(v); err == nil {
+		if d := time.Until(t); d > 0 {
+			return d
+		}
+	}
+	return 0
+}
+
+// retryAfterOfError extracts Retry-After from a typed transient error.
+func retryAfterOfError(err error) time.Duration {
+	var te *TransientError
+	if errors.As(err, &te) {
+		return te.RetryAfter
+	}
+	return 0
+}
+
+// backoffDelay returns the deterministic wait before retry `attempt`
+// (counting from 1): Base*2^(attempt-1), capped at Max.
+func backoffDelay(p RetryPolicy, attempt int) time.Duration {
+	if p.Base <= 0 {
+		return 0
+	}
+	wait := p.Base
+	for i := 1; i < attempt; i++ {
+		if p.Max > 0 && wait >= p.Max {
+			return p.Max
+		}
+		wait *= 2
+	}
+	if p.Max > 0 && wait > p.Max {
+		wait = p.Max
+	}
+	return wait
+}
+
 // DoRetry runs fn until it succeeds, fails non-transiently, or attempts run
-// out. Waits are deterministic: N*Base capped at Max, no jitter.
+// out. Waits are deterministic: Base*2^(N-1) capped at Max, no jitter. A
+// transient error's Retry-After wins when it is longer than the backoff, and
+// every wait arms the process-wide cooldown gate so concurrent and sequential
+// callers back off too.
 func DoRetry(ctx context.Context, p RetryPolicy, fn func() error) error {
 	if p.Attempts < 1 {
 		p.Attempts = 1
@@ -44,13 +155,18 @@ func DoRetry(ctx context.Context, p RetryPolicy, fn func() error) error {
 		if ctx.Err() != nil {
 			return ctx.Err()
 		}
-		if err = fn(); err == nil || !Transient(err) || attempt == p.Attempts {
+		err = fn()
+		if err == nil || !Transient(err) || attempt == p.Attempts {
 			return err
 		}
-		wait := time.Duration(attempt) * p.Base
-		if wait > p.Max {
-			wait = p.Max
+		wait := backoffDelay(p, attempt)
+		if ra := retryAfterOfError(err); ra > wait {
+			wait = ra
 		}
+		if wait <= 0 {
+			continue
+		}
+		globalCooldown.note(wait)
 		select {
 		case <-ctx.Done():
 			return ctx.Err()

retry_test.go

@@ -3,6 +3,8 @@ package onlyoffice
 import (
 	"context"
 	"errors"
+	"net/http"
+	"net/http/httptest"
 	"testing"
 	"time"
 )
@@ -20,6 +22,8 @@ func TestTransient(t *testing.T) {
 		{errors.New("PUT /x: 504 gateway timeout"), true},
 		{errors.New("GET /x: 404 not found"), false},
 		{errors.New("GET /x: 4290 not a code"), false},
+		{&TransientError{Code: 429, Msg: "throttled"}, true},
+		{&TransientError{Code: 404, Msg: "missing"}, false},
 	}
 	for _, c := range cases {
 		if got := Transient(c.err); got != c.want {
@@ -86,3 +90,131 @@ func TestDoRetryHonoursContextCancellation(t *testing.T) {
 		t.Fatalf("want context.Canceled, got %v", err)
 	}
 }
+
+func TestBackoffDelayExponentialAndCapped(t *testing.T) {
+	p := RetryPolicy{Base: 100 * time.Millisecond, Max: 250 * time.Millisecond}
+	want := []time.Duration{
+		100 * time.Millisecond, // 2^0
+		200 * time.Millisecond, // 2^1
+		250 * time.Millisecond, // 2^2 capped
+		250 * time.Millisecond,
+		250 * time.Millisecond,
+	}
+	for i, w := range want {
+		if got := backoffDelay(p, i+1); got != w {
+			t.Errorf("backoffDelay(attempt=%d) = %v, want %v", i+1, got, w)
+		}
+	}
+}
+
+func TestDoRetryWaitsRetryAfterLongerThanBackoff(t *testing.T) {
+	const ra = 60 * time.Millisecond
+	p := RetryPolicy{Attempts: 2, Base: time.Millisecond, Max: time.Millisecond}
+	attempts := 0
+	start := time.Now()
+	err := DoRetry(context.Background(), p, func() error {
+		attempts++
+		if attempts == 1 {
+			return &TransientError{Code: http.StatusTooManyRequests, RetryAfter: ra, Msg: "GET /x: 429"}
+		}
+		return nil
+	})
+	if err != nil {
+		t.Fatalf("DoRetry: %v", err)
+	}
+	if elapsed := time.Since(start); elapsed < ra {
+		t.Fatalf("waited %v, want >= %v", elapsed, ra)
+	}
+}
+
+func TestParseRetryAfter(t *testing.T) {
+	cases := []struct {
+		in   string
+		want time.Duration
+	}{
+		{"", 0},
+		{"  0 ", 0},
+		{"-3", 0},
+		{"bogus", 0},
+		{"120", 120 * time.Second},
+	}
+	for _, c := range cases {
+		if got := parseRetryAfter(c.in); got != c.want {
+			t.Errorf("parseRetryAfter(%q) = %v, want %v", c.in, got, c.want)
+		}
+	}
+	future := time.Now().Add(90 * time.Second).UTC().Format(http.TimeFormat)
+	if got := parseRetryAfter(future); got < 80*time.Second || got > 95*time.Second {
+		t.Errorf("HTTP-date Retry-After = %v, want ~90s", got)
+	}
+}
+
+// TestGetJSON429CarriesRetryAfter checks that the HTTP layer surfaces a 429 as
+// a typed transient error carrying the parsed Retry-After header.
+func TestGetJSON429CarriesRetryAfter(t *testing.T) {
+	t.Setenv("OO_RETRY_ATTEMPTS", "1")
+	resetPacingForTest()
+	defer resetPacingForTest()
+
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Retry-After", "1")
+		w.WriteHeader(http.StatusTooManyRequests)
+		_, _ = w.Write([]byte("busy"))
+	}))
+	defer srv.Close()
+
+	c := NewClient(Credentials{Url: srv.URL})
+	c.token = &Token{Value: "tok", Expires: Time(time.Now().Add(time.Hour))}
+
+	_, err := c.getJSON(context.Background(), "/api/2.0/project/4.json")
+	if err == nil {
+		t.Fatal("want error")
+	}
+	var te *TransientError
+	if !errors.As(err, &te) {
+		t.Fatalf("want *TransientError, got %T: %v", err, err)
+	}
+	if te.Code != http.StatusTooManyRequests {
+		t.Fatalf("code = %d, want 429", te.Code)
+	}
+	if te.RetryAfter != time.Second {
+		t.Fatalf("RetryAfter = %v, want 1s", te.RetryAfter)
+	}
+}
+
+// TestGetJSONRetries429ThenSucceeds verifies the end-to-end path: a throttled
+// answer with Retry-After is retried and the eventual success is returned.
+func TestGetJSONRetries429ThenSucceeds(t *testing.T) {
+	t.Setenv("OO_RETRY_ATTEMPTS", "3")
+	t.Setenv("OO_RETRY_BASE", "10ms")
+	t.Setenv("OO_RETRY_MAX", "50ms")
+	resetPacingForTest()
+	defer resetPacingForTest()
+
+	var calls int
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		calls++
+		if calls == 1 {
+			w.Header().Set("Retry-After", "0")
+			w.WriteHeader(http.StatusTooManyRequests)
+			return
+		}
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{"response":{}}`))
+	}))
+	defer srv.Close()
+
+	c := NewClient(Credentials{Url: srv.URL})
+	c.token = &Token{Value: "tok", Expires: Time(time.Now().Add(time.Hour))}
+
+	raw, err := c.getJSON(context.Background(), "/api/2.0/project/4.json")
+	if err != nil {
+		t.Fatalf("getJSON: %v", err)
+	}
+	if string(raw) != `{"response":{}}` {
+		t.Fatalf("raw = %s", raw)
+	}
+	if calls != 2 {
+		t.Fatalf("calls = %d, want 2", calls)
+	}
+}

storage_fallback.go

@@ -150,7 +150,7 @@ func (c *Client) downloadFileEntry(ctx context.Context, f *FileEntry, dst io.Wri
 			}
 			return 0, fmt.Errorf("GET viewUrl: %d (stale S3) and minio fallback: %w", resp.StatusCode, merr)
 		}
-		return 0, fmt.Errorf("GET viewUrl: %d %s", resp.StatusCode, truncate(string(b), 400))
+		return 0, statusError(resp.StatusCode, retryAfterOf(resp), "GET viewUrl: %d %s", resp.StatusCode, truncate(string(b), 400))
 	}
 	return io.Copy(dst, resp.Body)
 }