Copybara | 854996b | 2021-09-07 19:36:02 +0000 | [diff] [blame] | 1 | # Copyright 2016 The Chromium Authors. All rights reserved. |
| 2 | # Use of this source code is governed by a BSD-style |
| 3 | # license that can be found in the LICENSE file or at |
| 4 | # https://developers.google.com/open-source/licenses/bsd |
| 5 | |
| 6 | """A simple in-RAM cache with distributed invalidation. |
| 7 | |
| 8 | Here's how it works: |
| 9 | + Each frontend or backend job has one CacheManager which |
| 10 | owns a set of RamCache objects, which are basically dictionaries. |
| 11 | + Each job can put objects in its own local cache, and retrieve them. |
| 12 | + When an item is modified, the item at the corresponding cache key |
| 13 | is invalidated, which means two things: (a) it is dropped from the |
| 14 | local RAM cache, and (b) the key is written to the Invalidate table. |
| 15 | + On each incoming request, the job checks the Invalidate table for |
| 16 | any entries added since the last time that it checked. If it finds |
| 17 | any, it drops all RamCache entries for the corresponding key. |
| 18 | + There is also a cron task that truncates old Invalidate entries |
| 19 | when the table is too large. If a frontend job sees more than the |
| 20 | max Invalidate rows, it will drop everything from all caches, |
| 21 | because it does not know what it missed due to truncation. |
| 22 | + The special key 0 means to drop all cache entries. |
| 23 | |
| 24 | This approach makes jobs use cached values that are not stale at the |
| 25 | time that processing of each request begins. There is no guarantee that |
| 26 | an item will not be modified by some other job and that the cached entry |
| 27 | could become stale during the lifetime of that same request. |
| 28 | |
| 29 | TODO(jrobbins): Listener hook so that client code can register its own |
| 30 | handler for invalidation events. E.g., the sorting code has a cache that |
| 31 | is correctly invalidated on each issue change, but needs to be completely |
| 32 | dropped when a config is modified. |
| 33 | |
| 34 | TODO(jrobbins): If this part of the system becomes a bottleneck, consider |
| 35 | some optimizations: (a) splitting the table into multiple tables by |
| 36 | kind, or (b) sharding the table by cache_key. Or, maybe leverage memcache |
| 37 | to avoid even hitting the DB in the frequent case where nothing has changed. |
| 38 | """ |
| 39 | from __future__ import print_function |
| 40 | from __future__ import division |
| 41 | from __future__ import absolute_import |
| 42 | |
| 43 | import collections |
| 44 | import logging |
| 45 | |
| 46 | from framework import jsonfeed |
| 47 | from framework import sql |
| 48 | |
| 49 | |
| 50 | INVALIDATE_TABLE_NAME = 'Invalidate' |
| 51 | INVALIDATE_COLS = ['timestep', 'kind', 'cache_key'] |
| 52 | # Note: *_id invalidations should happen only when there's a change |
| 53 | # in one of the values used to look up the internal ID number. |
| 54 | # E.g. hotlist_id_2lc should only be invalidated when the hotlist |
| 55 | # name or owner changes. |
| 56 | INVALIDATE_KIND_VALUES = [ |
| 57 | 'user', 'usergroup', 'project', 'project_id', 'issue', 'issue_id', |
| 58 | 'hotlist', 'hotlist_id', 'comment', 'template' |
| 59 | ] |
| 60 | INVALIDATE_ALL_KEYS = 0 |
| 61 | MAX_INVALIDATE_ROWS_TO_CONSIDER = 1000 |
| 62 | |
| 63 | |
| 64 | class CacheManager(object): |
| 65 | """Service class to manage RAM caches and shared Invalidate table.""" |
| 66 | |
| 67 | def __init__(self): |
| 68 | self.cache_registry = collections.defaultdict(list) |
| 69 | self.processed_invalidations_up_to = 0 |
| 70 | self.invalidate_tbl = sql.SQLTableManager(INVALIDATE_TABLE_NAME) |
| 71 | |
| 72 | def RegisterCache(self, cache, kind): |
| 73 | """Register a cache to be notified of future invalidations.""" |
| 74 | assert kind in INVALIDATE_KIND_VALUES |
| 75 | self.cache_registry[kind].append(cache) |
| 76 | |
| 77 | def _InvalidateAllCaches(self): |
| 78 | """Invalidate all cache entries.""" |
| 79 | for cache_list in self.cache_registry.values(): |
| 80 | for cache in cache_list: |
| 81 | cache.LocalInvalidateAll() |
| 82 | |
| 83 | def _ProcessInvalidationRows(self, rows): |
| 84 | """Invalidate cache entries indicated by database rows.""" |
| 85 | already_done = set() |
| 86 | for timestep, kind, key in rows: |
| 87 | self.processed_invalidations_up_to = max( |
| 88 | self.processed_invalidations_up_to, timestep) |
| 89 | if (kind, key) in already_done: |
| 90 | continue |
| 91 | already_done.add((kind, key)) |
| 92 | for cache in self.cache_registry[kind]: |
| 93 | if key == INVALIDATE_ALL_KEYS: |
| 94 | cache.LocalInvalidateAll() |
| 95 | else: |
| 96 | cache.LocalInvalidate(key) |
| 97 | |
| 98 | def DoDistributedInvalidation(self, cnxn): |
| 99 | """Drop any cache entries that were invalidated by other jobs.""" |
| 100 | # Only consider a reasonable number of rows so that we can never |
| 101 | # get bogged down on this step. If there are too many rows to |
| 102 | # process, just invalidate all caches, and process the last group |
| 103 | # of rows to update processed_invalidations_up_to. |
| 104 | rows = self.invalidate_tbl.Select( |
| 105 | cnxn, cols=INVALIDATE_COLS, |
| 106 | where=[('timestep > %s', [self.processed_invalidations_up_to])], |
| 107 | order_by=[('timestep DESC', [])], |
| 108 | limit=MAX_INVALIDATE_ROWS_TO_CONSIDER) |
| 109 | |
| 110 | cnxn.Commit() |
| 111 | |
| 112 | if len(rows) == MAX_INVALIDATE_ROWS_TO_CONSIDER: |
| 113 | logging.info('Invaliditing all caches: there are too many invalidations') |
| 114 | self._InvalidateAllCaches() |
| 115 | |
| 116 | logging.info('Saw %d invalidation rows', len(rows)) |
| 117 | self._ProcessInvalidationRows(rows) |
| 118 | |
| 119 | def StoreInvalidateRows(self, cnxn, kind, keys): |
| 120 | """Store rows to let all jobs know to invalidate the given keys.""" |
| 121 | assert kind in INVALIDATE_KIND_VALUES |
| 122 | self.invalidate_tbl.InsertRows( |
| 123 | cnxn, ['kind', 'cache_key'], [(kind, key) for key in keys]) |
| 124 | |
| 125 | def StoreInvalidateAll(self, cnxn, kind): |
| 126 | """Store a value to tell all jobs to invalidate all items of this kind.""" |
| 127 | last_timestep = self.invalidate_tbl.InsertRow( |
| 128 | cnxn, kind=kind, cache_key=INVALIDATE_ALL_KEYS) |
| 129 | self.invalidate_tbl.Delete( |
| 130 | cnxn, kind=kind, where=[('timestep < %s', [last_timestep])]) |
| 131 | |
| 132 | |
| 133 | class RamCacheConsolidate(jsonfeed.InternalTask): |
| 134 | """Drop old Invalidate rows when there are too many of them.""" |
| 135 | |
| 136 | def HandleRequest(self, mr): |
| 137 | """Drop excessive rows in the Invalidate table and return some stats. |
| 138 | |
| 139 | Args: |
| 140 | mr: common information parsed from the HTTP request. |
| 141 | |
| 142 | Returns: |
| 143 | Results dictionary in JSON format. The stats are just for debugging, |
| 144 | they are not used by any other part of the system. |
| 145 | """ |
| 146 | tbl = self.services.cache_manager.invalidate_tbl |
| 147 | old_count = tbl.SelectValue(mr.cnxn, 'COUNT(*)') |
| 148 | |
| 149 | # Delete anything other than the last 1000 rows because we won't |
| 150 | # look at them anyway. If a job gets a request and sees 1000 new |
| 151 | # rows, it will drop all caches of all types, so it is as if there |
| 152 | # were INVALIDATE_ALL_KEYS entries. |
| 153 | if old_count > MAX_INVALIDATE_ROWS_TO_CONSIDER: |
| 154 | kept_timesteps = tbl.Select( |
| 155 | mr.cnxn, ['timestep'], |
| 156 | order_by=[('timestep DESC', [])], |
| 157 | limit=MAX_INVALIDATE_ROWS_TO_CONSIDER) |
| 158 | earliest_kept = kept_timesteps[-1][0] |
| 159 | tbl.Delete(mr.cnxn, where=[('timestep < %s', [earliest_kept])]) |
| 160 | |
| 161 | new_count = tbl.SelectValue(mr.cnxn, 'COUNT(*)') |
| 162 | |
| 163 | return { |
| 164 | 'old_count': old_count, |
| 165 | 'new_count': new_count, |
| 166 | } |