Extend retry backoff; rework logging (LOGLEVEL, request lines, colors)

Backoff now covers all dial-phase errors (refused, timeout, DNS), TLS handshake failures, and 5xx responses with replayable bodies; streamed POSTs are never replayed to avoid duplicate work. First retry of an episode logs at WARN, subsequent attempts at INFO. LOGLEVEL (LOG_LEVEL kept as alias) now defaults to warn: startup logs version plus every setting; INFO adds one line per incoming request and per response with status/duration, ANSI-colored in text mode (bypasses slog's escaping so colors render in docker compose logs); NO_COLOR or LOG_FORMAT=json disables colors.
2026-09-20 20:00:57 +02:00
parent 20d5439b5f
commit 42e1386811
5 changed files with 312 additions and 42 deletions
@@ -2,6 +2,7 @@ package proxy

 import (
 	"context"
+	"crypto/tls"
 	"io"
 	"net"
 	"net/http"
@@ -116,7 +117,91 @@ func TestRetryViaReverseProxy(t *testing.T) {
 	defer cancel()
 	req, _ := http.NewRequestWithContext(ctx, http.MethodGet, "http://127.0.0.1:1/", nil)
 	_, err = rt.RoundTrip(req)
-	if err == nil || !isConnRefused(err) {
+	if err == nil || !shouldRetry(err) {
 		t.Fatalf("err = %v, want connection refused", err)
 	}
 }
+
+// flakyStatus returns 500 for the first fails requests, then 200.
+type flakyStatus struct {
+	fails int
+	calls int
+}
+
+func (s *flakyStatus) RoundTrip(req *http.Request) (*http.Response, error) {
+	s.calls++
+	code := 200
+	if s.calls <= s.fails {
+		code = 500
+	}
+	return &http.Response{
+		StatusCode: code,
+		Status:     http.StatusText(code),
+		Body:       io.NopCloser(strings.NewReader("")),
+		Header:     make(http.Header),
+	}, nil
+}
+
+func TestRetryTransport5xxGet(t *testing.T) {
+	fs := &flakyStatus{fails: 2}
+	rt := &retryTransport{base: fs, initial: time.Millisecond, max: 2 * time.Millisecond}
+
+	req, _ := http.NewRequest(http.MethodGet, "http://upstream/api/version", nil)
+	resp, err := rt.RoundTrip(req)
+	if err != nil {
+		t.Fatal(err)
+	}
+	resp.Body.Close()
+	if resp.StatusCode != 200 {
+		t.Fatalf("status = %d, want 200", resp.StatusCode)
+	}
+	if fs.calls != 3 {
+		t.Fatalf("calls = %d, want 3", fs.calls)
+	}
+}
+
+func TestRetryTransport5xxPostNotRetried(t *testing.T) {
+	fs := &flakyStatus{fails: 10}
+	rt := &retryTransport{base: fs, initial: time.Millisecond, max: time.Millisecond}
+
+	// A streamed body (no GetBody) must not be replayed after a 500.
+	req, _ := http.NewRequest(http.MethodPost, "http://upstream/prompt", io.NopCloser(strings.NewReader("{}")))
+	resp, err := rt.RoundTrip(req)
+	if err != nil {
+		t.Fatal(err)
+	}
+	resp.Body.Close()
+	if resp.StatusCode != 500 {
+		t.Fatalf("status = %d, want 500", resp.StatusCode)
+	}
+	if fs.calls != 1 {
+		t.Fatalf("calls = %d, want 1 (no retry for streamed POST)", fs.calls)
+	}
+}
+
+func TestShouldRetryClassification(t *testing.T) {
+	cases := []struct {
+		name string
+		err  error
+		want bool
+	}{
+		{"dial refused", &net.OpError{Op: "dial", Net: "tcp", Err: syscall.ECONNREFUSED}, true},
+		{"dial timeout", &net.OpError{Op: "dial", Net: "tcp", Err: timeoutErr{}}, true},
+		{"dns", &net.DNSError{Err: "no such host", IsNotFound: true}, true},
+		{"tls record", tls.RecordHeaderError{Msg: "bad"}, true},
+		{"tls alert", tls.AlertError(42), true},
+		{"read error mid-request", &net.OpError{Op: "read", Net: "tcp", Err: syscall.ECONNRESET}, false},
+		{"plain error", io.EOF, false},
+	}
+	for _, c := range cases {
+		if got := shouldRetry(c.err); got != c.want {
+			t.Errorf("%s: shouldRetry = %v, want %v", c.name, got, c.want)
+		}
+	}
+}
+
+type timeoutErr struct{}
+
+func (timeoutErr) Error() string   { return "i/o timeout" }
+func (timeoutErr) Timeout() bool   { return true }
+func (timeoutErr) Temporary() bool { return true }