From efe21247534b1b8952af7a782aa456145e4a93d5 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Rapha=C3=ABl=20Vinot?= <raphael@vinot.info>
Date: Mon, 20 Nov 2023 11:45:41 +0100
Subject: [PATCH] fix: Quit BG indexer when shutdown is requested. Improve
 exceptions handling in archiver

---
 README.md                 | 68 +++++++++++++++++++++++++++++++++++++++
 bin/archiver.py           | 10 ++++--
 bin/background_indexer.py |  5 +++
 3 files changed, 80 insertions(+), 3 deletions(-)

diff --git a/README.md b/README.md
index 79bcf648..dd905cb3 100644
--- a/README.md
+++ b/README.md
@@ -65,6 +65,74 @@ pip install pylookyloo
 
 For more details on `pylookyloo`, read the overview [docs](https://www.lookyloo.eu/docs/main/pylookyloo-overview.html), the [documentation](https://pylookyloo.readthedocs.io/en/latest/) of the module itself, or the code in this [GitHub repository](https://github.com/Lookyloo/PyLookyloo).
 
+# Notes regarding using S3FS for storage
+
+## Directory listing
+
+TL;DR: it is slow.
+
+If you have namy captures (say more than 1000/day), and store captures in a s3fs bucket mounted with s3fs-fuse,
+doing a directory listing in bash (`ls`) will most probably lock the I/O for every process
+trying to access any file in the whole bucket. The same will be true if you access the
+filesystem using python methods (`iterdir`, `scandir`...))
+
+A workaround is to use the python s3fs module as it will not access the filesystem for listing directories.
+You can configure the s3fs credentials in `config/generic.json` key `s3fs`.
+
+## Versioning
+
+By default, a MinIO bucket (backend for s3fs) will have versioning enabled, wich means it
+keeps a copy of every version of every file you're storing. It becomes a problem if you have a lot of captures
+as the index files are updated on every change, and the max amount of versions is 10.000.
+So by the time you have > 10.000 captures in a directory, you'll get I/O errors when you try
+to update the index file. And you absolutely do not care about that versioning in lookyloo.
+
+To check if versioning is enabled (can be either enabled or suspended):
+
+```
+mc version info <alias_in_config>/<bucket>
+```
+
+The command below will suspend versioning:
+
+```bash
+mc version suspend <alias_in_config>/<bucket>
+```
+
+And if you're already stuck with an index that was updated 10.000 times and you cannot do anything about it:
+
+```bash
+mc rm --non-current --versions --recursive --force <alias_in_config>/<bucket>/path/to/index
+```
+
+Error message from bash (unhelpful):
+
+```bash
+$ (git::main) rm /path/to/lookyloo/archived_captures/Year/Month/Day/index
+rm: cannot remove '/path/to/lookyloo/archived_captures/Year/Month/Day/index': Input/output error
+```
+
+Python code:
+
+```python
+from lookyloo.default import get_config
+import s3fs
+
+s3fs_config = get_config('generic', 's3fs')
+s3fs_client = s3fs.S3FileSystem(key=s3fs_config['config']['key'],
+                                secret=s3fs_config['config']['secret'],
+                                endpoint_url=s3fs_config['config']['endpoint_url'])
+
+s3fs_bucket = s3fs_config['config']['bucket_name']
+s3fs_client.rm_file(s3fs_bucket + '/Year/Month/Day/index')
+```
+
+Error from python (somewhat more helpful):
+```
+OSError: [Errno 5] An error occurred (MaxVersionsExceeded) when calling the DeleteObject operation: You've exceeded the limit on the number of versions you can create on this object
+```
+
+
 # Contributing to Lookyloo
 To learn more about contributing to Lookyloo, see our [contributor guide](https://www.lookyloo.eu/docs/main/contributing.html).
 
diff --git a/bin/archiver.py b/bin/archiver.py
index 175ff08b..7b615495 100755
--- a/bin/archiver.py
+++ b/bin/archiver.py
@@ -324,9 +324,13 @@ class Archiver(AbstractManager):
             try:
                 new_capture_path = self.__archive_single_capture(capture_path)
                 capture_breakpoint -= 1
-            except OSError as e:
-                self.logger.warning(f'Unable to archive capture: {e}')
-            finally:
+            except OSError:
+                self.logger.exception(f'Unable to archive capture {capture_path}')
+                (capture_path / 'lock').unlink(missing_ok=True)
+            except Exception:
+                self.logger.exception(f'Critical exception while archiving {capture_path}')
+                (capture_path / 'lock').unlink(missing_ok=True)
+            else:
                 (new_capture_path / 'lock').unlink(missing_ok=True)
 
         if archiving_done:
diff --git a/bin/background_indexer.py b/bin/background_indexer.py
index 24a9e6ef..0630e37d 100755
--- a/bin/background_indexer.py
+++ b/bin/background_indexer.py
@@ -45,7 +45,12 @@ class BackgroundIndexer(AbstractManager):
         archive_interval = timedelta(days=get_config('generic', 'archive'))
         cut_time = (datetime.now() - archive_interval)
         for month_dir in make_dirs_list(self.lookyloo.capture_dir):
+            __counter_shutdown = 0
             for capture_time, path in sorted(get_sorted_captures_from_disk(month_dir, cut_time=cut_time, keep_more_recent=True), reverse=True):
+                __counter_shutdown += 1
+                if __counter_shutdown % 10 and self.shutdown_requested():
+                    self.logger.warning('Shutdown requested, breaking.')
+                    return False
                 if ((path / 'tree.pickle.gz').exists() or (path / 'tree.pickle').exists()):
                     # We already have a pickle file
                     self.logger.debug(f'{path} has a pickle.')