Follows-up f36b73e96c, which moved these classes to libs/objectcache.
* Fix wrong @throws in MemcachedPeclBagOStuff.
* Fix wrong @returns in MemcachedBagOStuff::getClient().
* Rename MWMemcached to MemcachedClient.
* Remove mention of 'memcached.php', which doesn't exist anymore.
Change-Id: I34dbc859be4778cea489fd2344f233f30452605c
Avoid having one wiki access another wiki's local keyspace.
Instead, use the global keyspace to share values across wikis.
Also, imitating wfMemcKey from wfForeignMemcKey was semantically
incorrect due to $wgCachePrefix having precedence. Most interfaces
(e.g. UserRightsProxy, FileRepo, JobQueue etc.) only have access
to the wiki id (dbname + prefix). The local cache configuration
for wgCachePrefix is not and shouldn't have to be exposed.
Start enforcing that local cache keys are left private and
to share keys, one must use global keys.
Global keys (prefixed with "global:") have their own space and we
can use the wiki-id as regular key segment for keys about users.
Also:
* Expose a method to keep formatting of this key in one place.
As it used used in many different places in core, as well
as in CentralAuth.
* Make use of wfWikiId() in getDefaultKeyspace() to avoid
duplicating this logic.
Change-Id: I58836a24b9e239f460ab489bd2fe8ced8259833c
This generally only effects wikis with no slave DBs,
but also matters if the master has non-zero LB load.
If the master ends up being used for DB_SLAVE, care
should be shown for cache-aside writes
Interesting WAN cache events are now logged.
Change-Id: I2cd8e84138263c13ea23beb9ab3d7562340e1fd3
Also removed a few MW dependencies from MemcachedPeclBagOStuff.
It still uses an IP class method, so it has to stay for now.
Change-Id: I8c5c83046c58fb58091d6ce11b2385208262460f
* Follow-up to 0c9fb12265
* Make sure colons actually get escaped
* Added more unit tests
* Also fixed the test actual/expected order
Change-Id: Ie04ea6059ee1eb6d1da8f30fefdec52fa49d38fb
For the sake of safety and correctness, the following BagOStuff::makeKey()
invocations should return distinct keys:
$cache->makeKey( 'ab:', 'cd' );
$cache->makeKey( 'ab', ':cd' );
That is not currently the case, because while we use ':' as a key path
separator, we don't escape ':' in the input supplied to makeKey(). So, make
BagOStuff::makeKeyInternal() URL-encode colons.
To prevent this from messing up the max. key length calculations, reproduce
this logic in MemcachedBagOStuff::makeKeyInternal(), in lieu of having the
method call its parent.
Change-Id: I83ea7e7336a1c9e64aa42284c2517089a736efe5
Currently we have to undo any transformations we apply to keys in getMulti() by
iterating over the response array keys reversing any changes. This is hairy and
complicated. So --
* Replace calls to MemcachedBagOStuff::encodeKey() with calls to a new method,
MemcachedBagOStuff::validateKeyEncoding(). The new method only validates that
the key is compatible with memcached. If it is not, it throws an exception.
If it is, it returns the key unmodified.
* Remove MemcachedBagOStuff::{encode,decode}Key(), since they no longer serve a
purpose, and have no callers.
Change-Id: If3e20c6a1a1b42fc1f2823aa660328e37c26eccb
The motivation for this patch came from seeing the following error in the
memcached log:
memcached ERROR: Memcached error for key "flowdb:flow_ref:wiki:by-source:\
v3:0:tawikinews:பூமிக்கு_அச்சுறுத்தலான_சிறுகோள்களைக்_கண்டுபிடிக்கும்_முயற்சியில்_தனியார்_நிறுவன0jம்:4.7" \
on server ":": A BAD KEY WAS PROVIDED/CHARACTERS OUT OF RANGE
I submitted a fix to Flow in I26e531f6, but I noticed that AbuseFilter had a
similar issue (fixed by Aaron in I27b51a4b), so I started thinking about how
to solve this more generally:
* The regular expression we current use to sanitize keys does not cover
characters outside the ASCII range, but such characters can be illegal
if one of their constituent bytes (when taken by itself) is an ASCII
control character. So change the regular expression to cover any and all
characters that fall outside the range \x22-\x7e (and '#' -- see below).
* Enforce a key length limit of 255 bytes, which is the maximum length
permitted by the memcached protocol. The Tamil segment in the key above is 84
characters, but 233 bytes in UTF-8, which become 684 characters when
URL-encoded. To fix this, try to shrink any segment that would push the total
key length over the limit by md5()ing it. If the end result is *still* over
the limit (this would happen if, for example, $args consists of many short
strings), then concatenate all args together and MD5 them.
* MD5'd arguments are prefixed with '#'. Any "organic" '#'s in the key segments
are URL-encoded.
Change-Id: Ia46987d3b0a09bb6b1952abd936d4c72ea7c56a0
* This blocks on writing to all replicas
and returns false if any failed.
* This is useful if ChronologyProtector is to work across
domains by having the writes go everywhere so that later
reads will see them (and be local at the same time).
* Redundant doc comments were also removed.
Change-Id: I9ed098d563c64dba605e7809bc96731da3b3e79d
Implement merge() by getting a lock on the tier 1
cache, getting it's value, generating the new one,
writing to all tiers, and then unlocking tier 1.
This is done by just using the stock mergeViaLock().
This means that if tier 1 is in the same DC, the
other tiers only need 1 request each, just for set().
Change-Id: I4b0c303ef3b86b63e7630032ed0b010e79706324
Inject the DeferredUpdates::addCallableUpdate method via the
ObjectCache. This brings it closer to being able to move to /libs.
Change-Id: Ifa0d893002c3d709a4dc7346c263a92162274bd7
* Add a string `keyspace` member to BagOStuff instances. The default
implementation, meant for simple key/value stores, treats the key space
as a string prefix to prepend to keys. By default, its value is `local`,
but any instance created via ObjectCache::newFromParams() (or or one of
its callers) will have that default to $wgCachePrefix / wfWikiID().
* Add `makeKey` and `makeGlobalKey` methods to the base BagOStuff class.
These methods are not static to allow for BagOStuff types which require
a configured instance to know the underlying storage engine's key semantics.
* Make wfMemcKey() and wfGlobalCacheKey() delegate to these methods on the main
ObjectCache instance.
Change-Id: Ib7fc2f939be3decfa97f66af8c2431c51039905f
Provide a complement to ObjectCache::getMainWANInstance() and
ObjectCache::getMainStashInstance() which gets the default ObjectCache
instance.
Change-Id: Ib16ee40908b159e60be41a308db49a9291b5de0d
* Callers of get() no longer have to contend with
the annoying $casToken parameter, which is there
but totally unusable to non-BagOStuff code.
* The default get() now delegates to doGet(),
which callers must implement instead. They can
ignore the overhead of generating $casToken if
they do not implement cas(), which applies to
callers that use the stock merge(). If cas() is
used for merge(), then getWithToken() must be
implemented.
* Also add BagOStuff::READ_LATEST to mergeViaCas()
for sanity, as that missing before.
Likewise with mergeViaLock().
Change-Id: I4efce6a9ab4b1eadd2f161dff641004a7239c516
* This lets multiwrite backends upgrade cached items
to higher tiers using UPGRADE_TTL.
* This is useful for memcached/sql tiers or apc/memcached.
Change-Id: I34b30ce8b54f8de36429d48c80f6768aed310272
* Simplify the debug log call and use queries group
* Remove $wgDebugDumpSqlLength, as profiler output
already has shortened query strings (one can use
profiling without DBO_DEBUG)
* Removed $wgDebugDBTransactions as BEGIN/COMMIT already show
* Removed PostgresTransactionState as it was only used for
$wgDebugDBTransactions handling
* This cuts down on lots of global variable usage
Change-Id: I185adb1694441d074dea965960429b4910727620
Because in PHP5.3:
php > $a = "hash";
php > echo isset( $a['fallback'] );
1
php > echo $a['fallback'];
h
This will fix using MediaWiki with MySQL on PHP5.3
if neither APC, xcache nor wincache are available.
Change-Id: Iebf034be75b282e2654cd298713455caf062eda4
Flow had a key: flowdb:flow_ref:wiki:by-source:v3:Parser\'s_"broken"_+_(page)_&_grill:testwiki:1:4.7
the '+' in there was not being encoded (it only does /[\x00-\x20\x25\x7f]+/)
but coming back, it was decoded into ' '.
getMulti() shows a key=>value array or results. Since key was different,
we couldn't find what we had requested.
Bug: T110326
Change-Id: Ia92edd73d0eb7fe0d35e38e7e7af8174fb85cbcc
MultiWriteBagOStuff::__construct() barfs if the 'caches' parameter is unset,
but it permits it to be an empty array. This means that subsequent operations
that need to reference the first (primary) cache backend need to check that
it is there at all.
Since there is not much sense in having a MultiWriteBagOStuff object with zero
cache backends, make the constructor require a nonempty array for 'caches', and
remove the subsequent checks that are made redundant by this new restriction.
Change-Id: I30f3e0dcbfe67570a368e64b8233cc0ba7f90b2f
* This is useful for ParserCache, as it tries to focus on memcached
and use other caches (e.g. mariadb) for the long-tail of less used
content, as setup on WMF. The class uses BagOStuff in a way that is
compatible with this approach.
Bug: T109751
Change-Id: Ia64eb44a9b52a988fde27b468d604d9163bed4b4
* Use the nx/ex flags for the redis SET method
to implement add() correctly. This also handes a
prior FIXME comment.
* Made merge() work via locking instead of cas()
since Redis::MULTI does not work with twemproxy.
Locking, which uses add(), does, and works better
than it did before.
* Removed some pointless newlines.
Change-Id: I652bd0ad2c594097d2cb1ab77f291e8bd27ad14f
Keys are sent to Memcached encoded. However, getMulti()
will respond in [key => value] format. The keys it
responds with should not be the encoded versions, or
callers won't be able to map them to the results.
Bug: T111138
Change-Id: I0d821b1219a492be8e93453f0249c78f18e24533
