Show More
@@ -24,6 +24,7 b' import logging' | |||
|
24 | 24 | import time |
|
25 | 25 | import typing |
|
26 | 26 | import zlib |
|
27 | import sqlite3 | |
|
27 | 28 | |
|
28 | 29 | from rhodecode.lib.ext_json import json |
|
29 | 30 | from .lock import GenerationLock |
@@ -38,6 +39,72 b' NO_VAL = -917' | |||
|
38 | 39 | MODE_BINARY = 'BINARY' |
|
39 | 40 | |
|
40 | 41 | |
|
42 | EVICTION_POLICY = { | |
|
43 | 'none': { | |
|
44 | 'evict': None, | |
|
45 | }, | |
|
46 | 'least-recently-stored': { | |
|
47 | 'evict': 'SELECT {fields} FROM archive_cache ORDER BY store_time', | |
|
48 | }, | |
|
49 | 'least-recently-used': { | |
|
50 | 'evict': 'SELECT {fields} FROM archive_cache ORDER BY access_time', | |
|
51 | }, | |
|
52 | 'least-frequently-used': { | |
|
53 | 'evict': 'SELECT {fields} FROM archive_cache ORDER BY access_count', | |
|
54 | }, | |
|
55 | } | |
|
56 | ||
|
57 | ||
|
58 | class DB: | |
|
59 | ||
|
60 | def __init__(self): | |
|
61 | self.connection = sqlite3.connect(':memory:') | |
|
62 | self._init_db() | |
|
63 | ||
|
64 | def _init_db(self): | |
|
65 | qry = ''' | |
|
66 | CREATE TABLE IF NOT EXISTS archive_cache ( | |
|
67 | rowid INTEGER PRIMARY KEY, | |
|
68 | key_file TEXT, | |
|
69 | key_file_path TEXT, | |
|
70 | filename TEXT, | |
|
71 | full_path TEXT, | |
|
72 | store_time REAL, | |
|
73 | access_time REAL, | |
|
74 | access_count INTEGER DEFAULT 0, | |
|
75 | size INTEGER DEFAULT 0 | |
|
76 | ) | |
|
77 | ''' | |
|
78 | ||
|
79 | self.sql(qry) | |
|
80 | self.connection.commit() | |
|
81 | ||
|
82 | @property | |
|
83 | def sql(self): | |
|
84 | return self.connection.execute | |
|
85 | ||
|
86 | def bulk_insert(self, rows): | |
|
87 | qry = ''' | |
|
88 | INSERT INTO archive_cache ( | |
|
89 | rowid, | |
|
90 | key_file, | |
|
91 | key_file_path, | |
|
92 | filename, | |
|
93 | full_path, | |
|
94 | store_time, | |
|
95 | access_time, | |
|
96 | access_count, | |
|
97 | size | |
|
98 | ) | |
|
99 | VALUES ( | |
|
100 | ?, ?, ?, ?, ?, ?, ?, ?, ? | |
|
101 | ) | |
|
102 | ''' | |
|
103 | cursor = self.connection.cursor() | |
|
104 | cursor.executemany(qry, rows) | |
|
105 | self.connection.commit() | |
|
106 | ||
|
107 | ||
|
41 | 108 | class FileSystemCache: |
|
42 | 109 | |
|
43 | 110 | def __init__(self, index, directory, **settings): |
@@ -225,9 +292,76 b' class FanoutCache:' | |||
|
225 | 292 | def __contains__(self, item): |
|
226 | 293 | return self.has_key(item) |
|
227 | 294 | |
|
228 | def evict(self): | |
|
229 | """Remove old items based on the conditions""" | |
|
230 | # TODO: Implement this... | |
|
295 | def evict(self, policy=None, size_limit=None): | |
|
296 | """ | |
|
297 | Remove old items based on the conditions | |
|
298 | ||
|
299 | ||
|
300 | explanation of this algo: | |
|
301 | iterate over each shard, then for each shard iterate over the .key files | |
|
302 | read the key files metadata stored. This gives us a full list of keys, cached_archived, their size and | |
|
303 | access data, time creation, and access counts. | |
|
304 | ||
|
305 | Store that into a memory DB so we can run different sorting strategies easily. | |
|
306 | Summing the size is a sum sql query. | |
|
307 | ||
|
308 | Then we run a sorting strategy based on eviction policy. | |
|
309 | We iterate over sorted keys, and remove each checking if we hit the overall limit. | |
|
310 | """ | |
|
311 | ||
|
312 | policy = policy or self._eviction_policy | |
|
313 | size_limit = size_limit or self._cache_size_limit | |
|
314 | ||
|
315 | select_policy = EVICTION_POLICY[policy]['evict'] | |
|
316 | ||
|
317 | if select_policy is None: | |
|
318 | return 0 | |
|
319 | ||
|
320 | db = DB() | |
|
321 | ||
|
322 | data = [] | |
|
323 | cnt = 1 | |
|
324 | for shard in self._shards: | |
|
325 | for key_file in os.listdir(shard._directory): | |
|
326 | if key_file.endswith('.key'): | |
|
327 | key_file_path = os.path.join(shard._directory, key_file) | |
|
328 | with open(key_file_path, 'rb') as f: | |
|
329 | metadata = json.loads(f.read()) | |
|
330 | # in case we don't have size re-calc it... | |
|
331 | if not metadata.get('size'): | |
|
332 | fn = metadata.get('full_path') | |
|
333 | size = os.stat(fn).st_size | |
|
334 | ||
|
335 | data.append([ | |
|
336 | cnt, | |
|
337 | key_file, | |
|
338 | key_file_path, | |
|
339 | metadata.get('filename'), | |
|
340 | metadata.get('full_path'), | |
|
341 | metadata.get('store_time', 0), | |
|
342 | metadata.get('access_time', 0), | |
|
343 | metadata.get('access_count', 0), | |
|
344 | metadata.get('size', size), | |
|
345 | ]) | |
|
346 | cnt += 1 | |
|
347 | ||
|
348 | # Insert bulk data using executemany | |
|
349 | db.bulk_insert(data) | |
|
350 | ||
|
351 | ((total_size,),) = db.sql('SELECT COALESCE(SUM(size), 0) FROM archive_cache').fetchall() | |
|
352 | ||
|
353 | select_policy_qry = select_policy.format(fields='key_file_path, full_path, size') | |
|
354 | sorted_keys = db.sql(select_policy_qry).fetchall() | |
|
355 | ||
|
356 | for key, cached_file, size in sorted_keys: | |
|
357 | # simulate removal impact BEFORE removal | |
|
358 | total_size -= size | |
|
359 | if total_size <= size_limit: | |
|
360 | # we obtained what we wanted... | |
|
361 | break | |
|
362 | ||
|
363 | os.remove(cached_file) | |
|
364 | os.remove(key) | |
|
231 | 365 | return |
|
232 | 366 | |
|
233 | 367 |
General Comments 0
You need to be logged in to leave comments.
Login now