Also implement Serializable in to add stability to anything that
uses naïve serialization (such as for persistent caching).
Change-Id: I03ba654ffd80ba027d47d5d7583abfe48d52818d
Proxy to the "master"/"write" cache object method. This is similar to the
approach taken in MultiWriteBagOStuff
Bug: T198279
Change-Id: If0933246b7ef4fc07ebeec4c3c9625b1137dbe05
Put the object/callback fields first, followed by config fields
and then state fields, similar to LBFactory.
Change-Id: I40ceaa48842b505f15efe5e41ca7a6de437b6ff1
Sqlite used the base implementation of trying a SELECT 1 query and
seeing if it failed. Instead, make it use the sqlite_master table.
Also remove the base version of that method since it would always
cause this problem and all subclasses have proper implementations.
Make LoadBalancerTest::assertWriteAllowed() more explicit and add
more assertions there.
Change-Id: I6c7b0bea8894c45dfe8931748d6687f0e5d1e101
* Provide a sensible default for the HTTP timeout, and make it configurable.
* Also make HTTP proxy and caBundlePath configurable.
* Pass through the logger object to the HTTP client, as in EtcdConfig.
* Do not treat a 404 response code for DELETE or 204 for PUT as errors. The
doc comment already said that delete() should return true if the item
was not found.
Change-Id: I6a90a25339c3a433381a5300c8c7c1a644e2a10f
This will usually let one distinguish asynchronous preemptive cache refreshes
from synchronous value recomputations due to actual cache misses.
Change-Id: I3424a89cc96a12fe283460eeaf722ee0cf299534
The "tideways" extension was renamed by their developers, see
https://tideways.com/profiler/blog/releasing-new-tideways-xhprof-extension
While doing so, they also renamed enable/disable functions:
tideways_enable -> tideways_xhprof_enable
tideways_disable -> tideways_xhprof_disable
Change-Id: I63bc97dba461dd46241a094dfc59439c0d28a219
Follows-up 7fdc3d09a3. Also document the getData method.
This method seems unused from a quick glance at Code Search, but
there's a lot of false positives from Monolog instances.
To play safe, rather than removing, add hard-deprecation for
a few days just in case we find another hit.
Change-Id: I8aab93744e66c1a583c67f0bbac27a45250a8fbb
* PHP 7 only checks for request/connection timeout after the timeout
passed to curl_multi_select() expires. So it's necessary to use a
select timeout which is much shorter than 10s. I filed
https://bugs.php.net/bug.php?id=76480 for this.
* Use millisecond resolution timeout parameters. These have been
available since libcurl 7.16.2, released Jan 2008, available in
Debian 7, Ubuntu 14.04 and CentOS 6.
Change-Id: Ia07b824dde179b33e14b81a76d580ce547bca315
Replace it all with random_bytes(), leave
only MWCryptRand::generateHex() as a convenience helper.
Change-Id: Ic30376a90e66d8f00dab86e7e6466fb3a750b87d
Use bisection to find the position within the ring, as in ketama.
According to my benchmarking this is faster than the guess/refine
algorithm for rings with 100 locations, i.e. 16000 ring nodes requiring
at most 14 bisection iterations.
Change-Id: I6042f382de093081971b5ec210eda8dfa74988f2
Uses new PHP 5.6 syntax like ...parameter unpacking and
calling anything looking like a callback to make the code more readable.
There are much more occurrences but this commit is intentionally limited
to an easily reviewable size.
In one occurrence, a simple conditional instead of trickery was much more readable.
This patch finishes all the easy stuf in the core, the remainder is either unobvious
or would result in smaller readability gains. It will be carefully dealt with in
further commits.
Change-Id: I79a16c48bfb98b75e5b99f2f6f4fa07b3ae02c5b
This scheme avoids disruptive re-hashing when a node is
ejected; only the adjectent clockwise sector is affected,
the later taking responsibility for the former. Addition
of new nodes is likewise less disruptive.
When used with MD5, this maps keys the same as libketama.
Also:
* Make HashRing::getLocations() faster by guessing the
sector list index and moving clockwise/counter-clockwise.
* Made HashRing Serializable for more robust storage.
* Allow for different hash functions to be specified.
Note:
* This makes PoolCounterRedis use the new hash scheme.
* Likewise for JobQueueFederated (for job de-duplication).
Switching from one mode to the other will cause some of
duplicated effort from old jobs only temporarily.
Change-Id: Iff432a4e5dc3fb40b2dda70fb3dbe9dda0b74933
In cases where we're operating on text data (and not binary data),
use e.g. "\u{00A0}" to refer directly to the Unicode character
'NO-BREAK SPACE' instead of "\xc2\xa0" to specify the bytes C2h A0h
(which correspond to the UTF-8 encoding of that character). This
makes it easier to look up those mysterious sequences, as not all
are as recognizable as the no-break space.
This is not enforced by PHP, but I think we should write those in
uppercase and zero-padded to at least four characters, like the
Unicode standard does.
Note that not all "\xNN" escapes can be automatically replaced:
* We can't use Unicode escapes for binary data that is not UTF-8
(e.g. in code converting from legacy encodings or testing the
handling of invalid UTF-8 byte sequences).
* '\xNN' escapes in regular expressions in single-quoted strings
are actually handled by PCRE and have to be dealt with carefully
(those regexps should probably be changed to use the /u modifier).
* "\xNN" referring to ASCII characters ("\x7F" and lower) should
probably be left as-is.
The replacements in this commit were done semi-manually by piping
the existing "\xNN" escapes through the following terrible Ruby
script I devised:
chars = eval('"' + ARGV[0] + '"').force_encoding('utf-8')
puts chars.split('').map{|char|
'\\u{' + char.ord.to_s(16).upcase.rjust(4, '0') + '}'
}.join('')
Change-Id: Idc3dee3a7fb5ebfaef395754d8859b18f1f8769a
The get() $cas_token parameter was changed into $flags, which can act
as a switch to make the return value an associative array of details.
pecl and PHP-based memcached BagOStuff class both pass all tests now.
Bug: T196125
Change-Id: I4678250331a48db4d50d1fc6c489c991a4dee920
Previously, if an internal service forwarded the cookies for a
user (e.g. for permissions) but not the User-Agent header or not
the IP address (e.g. XFF), ChronologyProtector could timeout
waiting for a matching writeIndex to appear for the wrong key.
The cookie now tethers the client to the key that holds the
DB positions from their last state-changing request.
Bug: T194403
Bug: T190082
Change-Id: I84f2cbea82532d911cdfed14644008894498813a
* Provide a default lock-based BagOStuff::cas implementation
* Make RedisBagOStuff::merge() use mergeViaCas()
* Use the raw unserialized string as the redis CAS token to
avoid any bad interaction with __wakeup() methods changing
field values every time
* Make RedisBagOStuff::incr() return false when there is no
such key, not null
* Rewrite merge() test to make the order of write/cas phase
of the parent and child merge() calls well defined instead
of arbitrary usleep() calls
* Avoid cache key reuse in test runs
Change-Id: I388ec173cf3858bb2fc7a8c8a00cda68703074ce
