Merge branch 'develop' into erikj/groups_merged

2 files changed, 190 insertions, 13 deletions

diff --git a/synapse/rest/media/v1/filepath.py b/synapse/rest/media/v1/filepath.py
index d92b7ff337..d5cec10127 100644
--- a/synapse/rest/media/v1/filepath.py
+++ b/synapse/rest/media/v1/filepath.py
@@ -14,6 +14,9 @@
 # limitations under the License.
 
 import os
+import re
+
+NEW_FORMAT_ID_RE = re.compile(r"^\d\d\d\d-\d\d-\d\d")
 
 
 class MediaFilePaths(object):
@@ -73,19 +76,105 @@ class MediaFilePaths(object):
         )
 
     def url_cache_filepath(self, media_id):
-        return os.path.join(
-            self.base_path, "url_cache",
-            media_id[0:2], media_id[2:4], media_id[4:]
-        )
+        if NEW_FORMAT_ID_RE.match(media_id):
+            # Media id is of the form <DATE><RANDOM_STRING>
+            # E.g.: 2017-09-28-fsdRDt24DS234dsf
+            return os.path.join(
+                self.base_path, "url_cache",
+                media_id[:10], media_id[11:]
+            )
+        else:
+            return os.path.join(
+                self.base_path, "url_cache",
+                media_id[0:2], media_id[2:4], media_id[4:],
+            )
+
+    def url_cache_filepath_dirs_to_delete(self, media_id):
+        "The dirs to try and remove if we delete the media_id file"
+        if NEW_FORMAT_ID_RE.match(media_id):
+            return [
+                os.path.join(
+                    self.base_path, "url_cache",
+                    media_id[:10],
+                ),
+            ]
+        else:
+            return [
+                os.path.join(
+                    self.base_path, "url_cache",
+                    media_id[0:2], media_id[2:4],
+                ),
+                os.path.join(
+                    self.base_path, "url_cache",
+                    media_id[0:2],
+                ),
+            ]
 
     def url_cache_thumbnail(self, media_id, width, height, content_type,
                             method):
+        # Media id is of the form <DATE><RANDOM_STRING>
+        # E.g.: 2017-09-28-fsdRDt24DS234dsf
+
         top_level_type, sub_type = content_type.split("/")
         file_name = "%i-%i-%s-%s-%s" % (
             width, height, top_level_type, sub_type, method
         )
-        return os.path.join(
-            self.base_path, "url_cache_thumbnails",
-            media_id[0:2], media_id[2:4], media_id[4:],
-            file_name
-        )
+
+        if NEW_FORMAT_ID_RE.match(media_id):
+            return os.path.join(
+                self.base_path, "url_cache_thumbnails",
+                media_id[:10], media_id[11:],
+                file_name
+            )
+        else:
+            return os.path.join(
+                self.base_path, "url_cache_thumbnails",
+                media_id[0:2], media_id[2:4], media_id[4:],
+                file_name
+            )
+
+    def url_cache_thumbnail_directory(self, media_id):
+        # Media id is of the form <DATE><RANDOM_STRING>
+        # E.g.: 2017-09-28-fsdRDt24DS234dsf
+
+        if NEW_FORMAT_ID_RE.match(media_id):
+            return os.path.join(
+                self.base_path, "url_cache_thumbnails",
+                media_id[:10], media_id[11:],
+            )
+        else:
+            return os.path.join(
+                self.base_path, "url_cache_thumbnails",
+                media_id[0:2], media_id[2:4], media_id[4:],
+            )
+
+    def url_cache_thumbnail_dirs_to_delete(self, media_id):
+        "The dirs to try and remove if we delete the media_id thumbnails"
+        # Media id is of the form <DATE><RANDOM_STRING>
+        # E.g.: 2017-09-28-fsdRDt24DS234dsf
+        if NEW_FORMAT_ID_RE.match(media_id):
+            return [
+                os.path.join(
+                    self.base_path, "url_cache_thumbnails",
+                    media_id[:10], media_id[11:],
+                ),
+                os.path.join(
+                    self.base_path, "url_cache_thumbnails",
+                    media_id[:10],
+                ),
+            ]
+        else:
+            return [
+                os.path.join(
+                    self.base_path, "url_cache_thumbnails",
+                    media_id[0:2], media_id[2:4], media_id[4:],
+                ),
+                os.path.join(
+                    self.base_path, "url_cache_thumbnails",
+                    media_id[0:2], media_id[2:4],
+                ),
+                os.path.join(
+                    self.base_path, "url_cache_thumbnails",
+                    media_id[0:2],
+                ),
+            ]
diff --git a/synapse/rest/media/v1/preview_url_resource.py b/synapse/rest/media/v1/preview_url_resource.py
index b81a336c5d..895b480d5c 100644
--- a/synapse/rest/media/v1/preview_url_resource.py
+++ b/synapse/rest/media/v1/preview_url_resource.py
@@ -36,6 +36,9 @@ import cgi
 import ujson as json
 import urlparse
 import itertools
+import datetime
+import errno
+import shutil
 
 import logging
 logger = logging.getLogger(__name__)
@@ -70,6 +73,10 @@ class PreviewUrlResource(Resource):
 
         self.downloads = {}
 
+        self._cleaner_loop = self.clock.looping_call(
+            self._expire_url_cache_data, 10 * 1000
+        )
+
     def render_GET(self, request):
         self._async_render_GET(request)
         return NOT_DONE_YET
@@ -130,7 +137,7 @@ class PreviewUrlResource(Resource):
         cache_result = yield self.store.get_url_cache(url, ts)
         if (
             cache_result and
-            cache_result["download_ts"] + cache_result["expires"] > ts and
+            cache_result["expires_ts"] > ts and
             cache_result["response_code"] / 100 == 2
         ):
             respond_with_json_bytes(
@@ -239,7 +246,7 @@ class PreviewUrlResource(Resource):
             url,
             media_info["response_code"],
             media_info["etag"],
-            media_info["expires"],
+            media_info["expires"] + media_info["created_ts"],
             json.dumps(og),
             media_info["filesystem_id"],
             media_info["created_ts"],
@@ -253,8 +260,7 @@ class PreviewUrlResource(Resource):
         # we're most likely being explicitly triggered by a human rather than a
         # bot, so are we really a robot?
 
-        # XXX: horrible duplication with base_resource's _download_remote_file()
-        file_id = random_string(24)
+        file_id = datetime.date.today().isoformat() + '_' + random_string(16)
 
         fname = self.filepaths.url_cache_filepath(file_id)
         self.media_repo._makedirs(fname)
@@ -328,6 +334,88 @@ class PreviewUrlResource(Resource):
             "etag": headers["ETag"][0] if "ETag" in headers else None,
         })
 
+    @defer.inlineCallbacks
+    def _expire_url_cache_data(self):
+        """Clean up expired url cache content, media and thumbnails.
+        """
+        now = self.clock.time_msec()
+
+        # First we delete expired url cache entries
+        media_ids = yield self.store.get_expired_url_cache(now)
+
+        removed_media = []
+        for media_id in media_ids:
+            fname = self.filepaths.url_cache_filepath(media_id)
+            try:
+                os.remove(fname)
+            except OSError as e:
+                # If the path doesn't exist, meh
+                if e.errno != errno.ENOENT:
+                    logger.warn("Failed to remove media: %r: %s", media_id, e)
+                    continue
+
+            removed_media.append(media_id)
+
+            try:
+                dirs = self.filepaths.url_cache_filepath_dirs_to_delete(media_id)
+                for dir in dirs:
+                    os.rmdir(dir)
+            except:
+                pass
+
+        yield self.store.delete_url_cache(removed_media)
+
+        if removed_media:
+            logger.info("Deleted %d entries from url cache", len(removed_media))
+
+        # Now we delete old images associated with the url cache.
+        # These may be cached for a bit on the client (i.e., they
+        # may have a room open with a preview url thing open).
+        # So we wait a couple of days before deleting, just in case.
+        expire_before = now - 2 * 24 * 60 * 60 * 1000
+        media_ids = yield self.store.get_url_cache_media_before(expire_before)
+
+        removed_media = []
+        for media_id in media_ids:
+            fname = self.filepaths.url_cache_filepath(media_id)
+            try:
+                os.remove(fname)
+            except OSError as e:
+                # If the path doesn't exist, meh
+                if e.errno != errno.ENOENT:
+                    logger.warn("Failed to remove media: %r: %s", media_id, e)
+                    continue
+
+            try:
+                dirs = self.filepaths.url_cache_filepath_dirs_to_delete(media_id)
+                for dir in dirs:
+                    os.rmdir(dir)
+            except:
+                pass
+
+            thumbnail_dir = self.filepaths.url_cache_thumbnail_directory(media_id)
+            try:
+                shutil.rmtree(thumbnail_dir)
+            except OSError as e:
+                # If the path doesn't exist, meh
+                if e.errno != errno.ENOENT:
+                    logger.warn("Failed to remove media: %r: %s", media_id, e)
+                    continue
+
+            removed_media.append(media_id)
+
+            try:
+                dirs = self.filepaths.url_cache_thumbnail_dirs_to_delete(media_id)
+                for dir in dirs:
+                    os.rmdir(dir)
+            except:
+                pass
+
+        yield self.store.delete_url_cache_media(removed_media)
+
+        if removed_media:
+            logger.info("Deleted %d media from url cache", len(removed_media))
+
 
 def decode_and_calc_og(body, media_uri, request_encoding=None):
     from lxml import etree