fix: 페이스북 릴 수집 경로 보강

src/main/java/com/hlab/yanalyst/domain/social/FacebookPageFeedClient.java

@@ -32,6 +32,8 @@ public class FacebookPageFeedClient implements SocialFeedClient {
     /** Facebook가 데이터센터별로 빈 피드를 돌려주는 경우가 있어 공식 로케일 호스트를 순서대로 시도한다. */
     private static final List<String> RENDERER_HOSTS = List.of(
             "www.facebook.com", "ko-kr.facebook.com", "en-gb.facebook.com");
+    private static final long READER_RECENT_CACHE_SECONDS = 3_600;
+    private static final long READER_BASELINE_CACHE_SECONDS = 2_592_000;
 
     private final RestTemplate restTemplate;
     private final ObjectMapper objectMapper;
@@ -60,27 +62,30 @@ public class FacebookPageFeedClient implements SocialFeedClient {
             }
         }
         if (readerFallbackEnabled) {
+            for (ReaderAttempt attempt : readerAttempts(channel.getLastCollectedAt() == null)) {
                 try {
-                SocialFeedBatch batch = fetchViaReader(channel, false);
+                    SocialFeedBatch batch = fetchViaReader(channel, attempt);
-                if (batch.videos().isEmpty()) batch = fetchViaReader(channel, true);
                     if (!batch.videos().isEmpty()) return batch;
+                    lastFailure = new IllegalStateException("Facebook Reader 공개 피드에 릴이 없습니다.");
                 } catch (IllegalStateException e) {
                     lastFailure = e;
                 }
             }
+        }
         throw new IllegalStateException("Facebook 공개 피드를 읽지 못했습니다: "
                 + (lastFailure == null ? "빈 응답" : friendly(lastFailure)), lastFailure);
     }
 
     /** 운영 서버 IP가 Facebook에서 차단될 때 공개 페이지를 브라우저로 읽어 주는 Reader API를 사용한다. */
-    private SocialFeedBatch fetchViaReader(SocialChannel channel, boolean noCache) {
+    private SocialFeedBatch fetchViaReader(SocialChannel channel, ReaderAttempt attempt) {
-        String source = channel.getSourceUrl().replaceAll("/+$", "") + "/reels/";
+        URI uri = readerUri(channel.getSourceUrl(), attempt.sourceScheme());
-        URI uri = URI.create("https://r.jina.ai/" + source);
         HttpHeaders headers = new HttpHeaders();
         headers.set(HttpHeaders.USER_AGENT, "Mozilla/5.0");
+        headers.setAccept(List.of(MediaType.TEXT_PLAIN));
+        headers.setAcceptLanguageAsLocales(List.of(Locale.ENGLISH));
         headers.set("X-Engine", "browser");
-        if (noCache) headers.set("X-No-Cache", "true");
+        if (attempt.noCache()) headers.set("X-No-Cache", "true");
-        else headers.set("X-Cache-Tolerance", "3600");
+        else headers.set("X-Cache-Tolerance", Long.toString(attempt.cacheToleranceSeconds()));
         try {
             ResponseEntity<String> response = restTemplate.exchange(uri, HttpMethod.GET,
                     new HttpEntity<>(headers), String.class);
@@ -94,6 +99,24 @@ public class FacebookPageFeedClient implements SocialFeedClient {
         }
     }
 
+    static URI readerUri(String channelUrl, String sourceScheme) {
+        String source = channelUrl.replaceFirst("^https?", sourceScheme).replaceAll("/+$", "") + "/reels/";
+        return URI.create("https://r.jina.ai/" + source);
+    }
+
+    static List<ReaderAttempt> readerAttempts(boolean baselineCollection) {
+        long fallbackTolerance = baselineCollection
+                ? READER_BASELINE_CACHE_SECONDS
+                : READER_RECENT_CACHE_SECONDS;
+        return List.of(
+                new ReaderAttempt("https", true, 0),
+                new ReaderAttempt("http", true, 0),
+                new ReaderAttempt("https", false, READER_RECENT_CACHE_SECONDS),
+                new ReaderAttempt("http", false, fallbackTolerance));
+    }
+
+    record ReaderAttempt(String sourceScheme, boolean noCache, long cacheToleranceSeconds) { }
+
     private SocialFeedBatch fetchFromHost(SocialChannel channel, String host) {
         try {
             Map<String, Object> config = new LinkedHashMap<>();

src/test/java/com/hlab/yanalyst/domain/social/FacebookPageFeedClientTest.java

@@ -2,6 +2,8 @@ package com.hlab.yanalyst.domain.social;
 
 import org.junit.jupiter.api.Test;
 
+import java.net.URI;
+
 import static org.assertj.core.api.Assertions.assertThat;
 
 class FacebookPageFeedClientTest {
@@ -50,4 +52,18 @@ class FacebookPageFeedClientTest {
                 .containsExactly("2234599224051571", "926151600080380");
         assertThat(batch.videos().get(0).thumbnailUrl()).isEqualTo("https://cdn.example/one.jpg?x=1&y=2");
     }
+
+    @Test
+    void retriesReaderWithBothSourceSchemesAndUsesLongCacheOnlyForBaseline() {
+        assertThat(FacebookPageFeedClient.readerUri("https://www.facebook.com/phutunet/", "http"))
+                .isEqualTo(URI.create("https://r.jina.ai/http://www.facebook.com/phutunet/reels/"));
+
+        assertThat(FacebookPageFeedClient.readerAttempts(true))
+                .extracting(FacebookPageFeedClient.ReaderAttempt::sourceScheme)
+                .containsExactly("https", "http", "https", "http");
+        assertThat(FacebookPageFeedClient.readerAttempts(true).get(3).cacheToleranceSeconds())
+                .isEqualTo(2_592_000);
+        assertThat(FacebookPageFeedClient.readerAttempts(false).get(3).cacheToleranceSeconds())
+                .isEqualTo(3_600);
+    }
 }