fix: DNS precheck and retry/backoff for FB API calls; improved DNS logging

app.py

@@ -7,6 +7,7 @@ import time
 import gradio as gr
 import threading
 import requests
+import socket
 from dotenv import load_dotenv
 
 load_dotenv()
@@ -151,6 +152,20 @@ if __name__ == "__main__":
         if not fb_token or not fb_page:
             logger.error("Facebook token or page id missing")
             ok = False
+        else:
+            # quick DNS resolution check to avoid noisy stack traces when DNS fails
+            try:
+                socket.getaddrinfo("graph.facebook.com", 443)
+            except Exception as e:
+                logger.error("DNS resolution failed for graph.facebook.com: %s", e)
+                try:
+                    with open("log.txt", "a", encoding="utf-8") as lf:
+                        lf.write(f"[{__import__('time').strftime('%Y-%m-%d %H:%M:%S')}] FB_DNS_RESOLUTION_FAILED {e}\n")
+                except Exception:
+                    logger.exception("Failed to write DNS failure to log.txt")
+                ok = False
+                # skip further FB checks when DNS is not available
+                return ok
         else:
             # Try to inspect the token using the Graph API debug_token endpoint when possible
             # This provides detailed info (is_valid, expires_at, type) when app credentials are available
@@ -221,6 +236,15 @@ if __name__ == "__main__":
                                 lf.write(entry)
                         except Exception:
                             logger.exception("Failed to write token validation ok to log.txt")
+            except requests.exceptions.RequestException as e:
+                # network-related errors (including DNS/name resolution) are common in Spaces intermittently
+                logger.exception("Facebook token validation network failure: %s", e)
+                try:
+                    with open("log.txt", "a", encoding="utf-8") as lf:
+                        lf.write(f"[{__import__('time').strftime('%Y-%m-%d %H:%M:%S')}] FB_TOKEN_VALIDATION_NETWORK_EXCEPTION {e}\n")
+                except Exception:
+                    logger.exception("Failed to write token validation exception to log.txt")
+                ok = False
             except Exception as e:
                 logger.exception("Facebook token validation failed with exception")
                 try:

post_to_facebook.py

@@ -2,6 +2,9 @@ import os
 import argparse
 import logging
 import requests
+import socket
+import time
+from requests.exceptions import RequestException
 from dotenv import load_dotenv
 
 logging.basicConfig(level=logging.INFO, format="%(asctime)s %(levelname)s %(message)s")
@@ -22,13 +25,34 @@ def post_to_facebook(page_id: str, access_token: str, message: str, link: str |
     if link:
         payload["link"] = link
     logger.info("Posting to Facebook page %s", page_id)
-    resp = requests.post(url, data=payload)
+    # Ensure graph.facebook.com resolves before attempting to POST to avoid long, noisy stack traces
     try:
-        resp.raise_for_status()
-    except requests.HTTPError:
-        logger.error("Facebook API error: %s", resp.text)
-        _append_log(f"[{__import__('time').strftime('%Y-%m-%d %H:%M:%S')}] FB_POST_ERROR page={page_id} message={message[:120]!r} response={resp.text}")
+        socket.getaddrinfo("graph.facebook.com", 443)
+    except Exception as e:
+        logger.error("DNS resolution failed for graph.facebook.com: %s", e)
+        _append_log(f"[{__import__('time').strftime('%Y-%m-%d %H:%M:%S')}] FB_DNS_RESOLUTION_FAILED {e}")
         raise
+
+    # Retry with exponential backoff for transient network issues
+    attempts = 4
+    backoff = 2
+    last_exc = None
+    for attempt in range(1, attempts + 1):
+        try:
+            resp = requests.post(url, data=payload, timeout=15)
+            resp.raise_for_status()
+            last_exc = None
+            break
+        except RequestException as e:
+            last_exc = e
+            logger.warning("Attempt %s/%s: Facebook POST failed: %s", attempt, attempts, e)
+            if attempt < attempts:
+                time.sleep(backoff)
+                backoff *= 2
+            else:
+                logger.error("All Facebook POST attempts failed")
+                _append_log(f"[{__import__('time').strftime('%Y-%m-%d %H:%M:%S')}] FB_POST_ERROR page={page_id} message={message[:120]!r} exception={e}")
+                raise
     data = resp.json()
     logger.info("Post successful: %s", data)
     # append a concise entry to log.txt for auditing