By default the kafka implementation we use doesn't require any kind of
acknowledgment, it just throws messages into the wind and lets them sit
where they may. Implement an option for KafkaHandler to specify the
number of acks (number of replicas that must record the message) and
some error handling to throw exceptions as necessary when there is a
problem.
Bug: T135159
Change-Id: I859dc791072db407f908b2f36be0d6704f1a6256
When MediaWiki encounters an unhandled exception, the error message it produces
includes a randomly-generated token, which allows the exception details to be
looked up in the error logs. This is useful but narrow: would it not be useful
to have the ability to retrieve all log records associated with a particular
request, rather than just exception details? (Hint: yes.)
So: introduce the notion of a request-global unique ID, retrievable via
WebRequest::getRequestId(). When MediaWiki is behind Apache + mod_unique_id
(which provides the same facility) or some other software which sets a
UNIQUE_ID envvar, the value of that envvar is used as the request ID.
Otherwise, it is a randomly-generated 24-character string.
The request ID supplants exception-specific IDs; MWExceptionHandler::getLogId()
is deprecated, accordingly. The request ID is also added as an annotation to
all Monolog-processed log records, and is exposed client-side as 'wgRequestId'.
This allows developers to associate a page view with log records even when the
page view does not result in an unhandled exception. (For the WMF, I also
intend to add it as an annotation to profiling data).
The request ID is not a tracking token; it does not persist, and it is
associated with a backend request, not with a particular user or a particular
session. Like the data in the NewPP report, the request ID is designed to be
cacheable, so that if, for example, a developer notices something weird in the
HTML, s/he can associate the output with a backend request regardless of
whether the response was served from the cache or directly from the backend.
Some prior art:
* https://httpd.apache.org/docs/2.4/mod/mod_unique_id.html
* http://api.rubyonrails.org/classes/ActionDispatch/RequestId.html
* https://github.com/dabapps/django-log-request-id
* https://packagist.org/packages/php-middleware/request-id
* https://github.com/rhyselsmore/flask-request-id
Change-Id: Iaf90c20c330e0470b9b98627a0228cadefd301d1
Add a 'mwversion' value to log events processed by WikiProcessor to help
with classifying log events emitted by a wiki farm running mixed
versions of MediaWiki.
Bug: T125707
Change-Id: I5f5dfd051ebf251bec0d6e6a83a15e81f59540f3
Respect the "private" context variable for debug log events when passing
events from MediaWiki\Logger\LegacyLogger to MWDebug::debugMsg. Passing
debug log events marked as private to MWDebug was a regression
introduced by the PSR-3 logging system.
Restore handling of $dest argument to wfDebug which was removed in 1.25
with the PSR-3 logging conversion. The documentation for $dest = 'log'
has also been removed. This third debug log event handling option was
not implemented in the PSR-3 logging conversion in 1.25. A follow up
change will remove known usage of $dest = 'log' in core.
Bug: T122644
Change-Id: Ib1d999b8b54e584e3944b46e9163a700f11c2e72
Deployed usages of avro now use the new format, allowing
this old style will only cause pain to those that attempt
to use it in the future.
Change-Id: Id27ed7b529153ef7983b9fc3c1250b19dc4a516f
This ensures that, in case "composer install" has not been run,
the user will see the error message about setting up dependencies
(as opposed a plain "Class not found" error because some other
dependency was used first).
Change-Id: Ib6026123770d21cc9f8960a1de361c8178b1b044
Proposed upstream changes to Monolog\Logger will require an ability to
call setter methods on newly created Logger instances to tune them for
use in high volume logging situations. Adding support for making these
types of adjustments to MediaWiki's Monolog integration will benefit
users of the updated library when they land upstream.
Arbitrary setters are called by adding a 'calls' array to the logger
specification. This array uses method names as keys and method arguments
as values. This syntax mirrors the setter invocation behavior of
ObjectFactory with the notable omission of Closure expansion in the
argument list before calling the Logger's setter method.
Bug: T116550
Change-Id: I990c7f00f57451f14954542f5404491b2660a0b7
LoggerFactory::getInstance() will be called many times during the course
of handling a typical MediaWiki request. The interface_exists() guard
condition it uses is an attempt to provide an informative error message
when Composer managed libraries are not installed. This check is only
needed on the first invocation of getInstance() to be effective. Using
an additional boolean to guard the interface_exists() call will allow
the PHP runtime to avoid a potentially expensive (at least compared to
a static boolean comparison) function call.
This is the sort of thing that smells of premature optimization, but its
addition is in fact informed by examination of performance reports from
the Wikimedia production environment.
Bug: T115729
Change-Id: I437bcb5326b06145081f2b86f6c4d0c8dc1a318c
Doxygen doesn't quite grok PHP's use of the reverse solidus (backslash)
character as a namespace separator. The C++ based parser it uses needs
them to be escaped in comments just as if they were being used in
a literal string context in PHP.
Change-Id: I9aff9dd0fb74a95039da1091c2f247cf71fd085a
- Removed space after cast
- Removed spaces in array index
- Removed double spaces
- Added spaces around string concat
- Fixed mixed tabs and spaces at begin of line
Change-Id: I38e849723f055d2d4c05cba72f5c245a28e8d5da
This allows a logging channel to be configured to write
directly to kafka. Logs can be serialized either to json
blobs or the more compact apache avro format.
The Kafka handler for monolog needs a list of one of more
kafka servers to query cluster metadata from. This should be
able to use any monolog formatter, although some like
JsonFormatter require you to disable formatBatch as Kafka
protocol would prefer to encode each record independently in
the protocol. This requires the nmred/kafka-php library,
version >= 1.3.0.
Adds a new formatter which serializes to the apache avro
format. This is a compact binary format which uses pre-
defined schemas. This initial implementation is very simple
and takes the plain schemas as a constructor argument.
Adds a new option to MonologSpi to wrap handlers in a
BufferHandler. This doesn't flush until the request shuts
down and prevents any network requests in the logger from
adding latency to web requests.
Related mediawiki/vendor update: Ibfe4bd2036ae8e998e2973f07bd9a6f057691578
The necessary config is something like:
array(
'loggers' => array(
'CirrusSearchRequests' => array(
'handlers' => array( 'kafka' ),
),
),
'handlers' => array(
'kafka' => array(
'factory' => '\\MediaWiki\\Logger\\Monolog\\KafkaHandler::factory',
'args' => array( 'localhost:9092' ),
'formatter' => 'avro',
'buffer' => true,
),
),
'formatters' => array(
'avro' => array(
'class' => '\\MediaWiki\\Logger\\Monolog\\AvroFormatter',
'args' => array(
array(
'CirrusSearchRequests' => array(
'type' => 'record',
'name' => 'CirrusSearchRequests'
'fields' => array( ... )
),
),
),
),
),
)
Bug: T106256
Change-Id: I6ee744b3e5306af0bed70811b558a543eed22840
Make MediaWiki\Logger\LegacyLogger and
MediaWiki\Logger\Monolog\LineFormatter better at outputting stacktrace
information and provide support for 'exception' data in the logging
context that is a structured array in addition to the default Exception
object support. This works with MWExceptionHandler::handleFatalError
generated data that is provided by an HHVM interpreter and cannot be
delivered as an Exception class.
With this patch, a good value for LineFormatter's format would be:
"%datetime% %extra.host% %extra.wiki% %channel% %level_name%: %message%
%context% %exception%\n"
Bug: T89169
Bug: T107440
Change-Id: Ida01ed51c573e1654346e716723e543a1be63090
Add a Monolog Formatter class that uses
MWExceptionHandler::getRedactedTraceAsString when outputting stack
traces.
Bug: T107440
Change-Id: Ic580c137e27aac95435f7b073a18cf61820b172f
Follows-up 77a397125f. Also add unit test that would've caught
this "PHP Notice: Undefined variable: data" error.
Change-Id: I8a3bd9c8b685c2aa7a466e3d3c61ffa027be02fa
Replace wfDebugLog() calls in MWExceptionHandler with direct use of
LoggerFactory and LoggerInterface. Logged exceptions are added to the
log message context.
LegacyLogger is also updated to append stack traces to any log event
when $wgLogExceptionBacktrace is true and the PSR-3 recommendation of
passing the exception as an 'exception' context item.
Handling of context data in LegacyLogger is expanded to support arrays,
exceptions and common object types.
Bug: T88649
Change-Id: I71499d895582bdea033a2516c902e23e38084080
Allow post-initialization configuration of MonologSpi by providing
a `mergeConfig()` method that can be used to merge a given collection of
configuration data with the existing configuration.
Bug: T104584
Change-Id: Iba6f115a79dbc0060f64a9095467d147cf53b8ae
wfSuppressWarnings() and wfRestoreWarnings() were split out into a
separate library. All usages in core were replaced with the new
functions, and the wf* global functions are marked as deprecated.
Additionally, some uses of @ were replaced due to composer's autoloader
being loaded even earlier.
Ie1234f8c12693408de9b94bf6f84480a90bd4f8e adds the library to
mediawiki/vendor.
Bug: T100923
Change-Id: I5c35079a0a656180852be0ae6b1262d40f6534c4
Seen in fatal.log in production: ErrorException from line 264 of
/srv/mediawiki/php-1.26wmf1/includes/exception/MWExceptionHandler.php:
Fatal Error: Class undefined: MediaWiki\Logger\DateTimeZone
Bug: T95727
Change-Id: Icb303314caaef47ac767fbf593e92d09e818f147
Move the non-namespaced classes referencing external Monolog classes to
an isolated PHP source file so that they aren't brought into scope by
the need to load another shim class.
Bug: T95220
Change-Id: I8270b8d5cd25db5a0f84fa94f59a6555052ae1ae
Move the MWLogger PSR-3 logging related classes into the
MediaWiki\Logger namespace. Create shim classes to ease migration of
existing MWLoggerFactory usage to the namespaced classes.
Bug: T93406
Change-Id: I359cc81fbd2dcf8937742311dcc7d3dee08747b0
The Monolog\Handler\SamlingHandler class available since Monolog v1.12.0
is an upstreamed equivalent of MWLoggerMonologSamplingHandler.
Requires: I8790da95fd658234e35b2d846af35993ebcd80e9
Change-Id: I3841cbab95382a66098d90f5570fa0bf3521578a
Make MWLoggerMonologHandler replace a prefix of "{channel}" with the log
event channel when formatting an event for a udp2log stream. This allows
a simpler configuration for a typical Monolog configuration where
a separate file is desired for each logging channel that is emitted to
udp2log. One handler can be shared by multiple channels.
Change-Id: Ie2b24ba02b281b8d8cf2ad58d28874c710e1e2e5
Monolog 1.12.0 "fixed" Handler::isHandling() so that instead of a full
log record it is only passed an array with the log event's level.
MWLoggerMonologHandler was relying on a full record to allow inspecting
the channel name and looking for a 'private' flag in the context
information.
Update MWLoggerMonologHandler to do legacy processing checks in
Handler::write() where the full log event is present for inspection.
Bug: T89313
Change-Id: Ia878c2cb6bff47d6b35ff38ba3b7ac2ea5556565
The stock Monolog\Handler\SyslogUdpHandler only works with a specifically
tailored Formatter class and it's output cannot be fed directly into
Logstash's syslog input. This handler implements RFC 3164 in a way that
can be used with any Formatter and parsed directly by Logstash.
Bug: T88870
Change-Id: Ib098d5cb8fe9643742360bf51b54fc0e27996c0f
MWLogger was renamed MWLoggerFactory and reduced to a static utility
class in Ie474676. Reintroduce an MWLogger that proxies the public
static methods of MWLoggerFactory to ease the transition of users who
have implemented $wgMWLoggerDefaultSpi configurations that reference
MWLogger.
As noted in the class, this is a temporary bandaid that should be ripped
off before 1.25 becomes an official release.
Change-Id: Iaccb78a510c60aab2ff20a9aa7c0869699657388
Time wounds all heels. During the code review for the PSR-3 logging
introduction, several people asked me why we needed a wrapper
for Psr\Log\LoggerInterface if the point was to use the standard. At the
time I was convinced that it would be better to introduce the dependency
via a wrapper class so that we could use the wrapper to patch over any
deficiencies that we might find in the PSR-3 API. After going on to work
on a project to disentangle other MediaWiki components from internal
project dependencies I have suddenly and clearly seen the error of my
ways.
We still need a logger factory as PSR-3 does not specify
a standard mechanism for creating Psr\Log\LoggerInterface instances. My
solution is to convert MWLogger into MWLoggerFactory to retain a static
factory interface for creating PSR-3 loggers but remove the MWLogger
wrapper class itself in favor of direct exposure of
Psr\Log\LoggerInterface to the MediaWiki consumer classes.
Change-Id: Ie47467657dcf341991ada00827dca5e8eff95438
For some log groups, we only want to log them if they meet a certain
level of severity. An example of this is the current 'memcached-serious'
log group, which can be merged with the normal 'memcached' group
in the future, and report at a severity of ERROR.
This adds a 'level' parameter to the $wgDebugLogGroups, for example:
$wgDebugLogGroups['memcached'] = array(
'destination' => '/var/log/mw/memcached.log',
'level' => \Psr\Log\LogLevel::ERROR,
);
Bug: T85073
Change-Id: Ic53bc4c8e318ed188fe6f4e838e6789b3c3fd574
