netlogix / sentry
sentry client for Flow & Neos
Installs: 117 015
Dependents: 0
Suggesters: 1
Security: 0
Stars: 2
Watchers: 5
Forks: 1
Open Issues: 2
Type:neos-package
Requires
- php: ^7.4 || ^8.0
- ext-json: *
- ext-openssl: *
- neos/flow: ^7.3.6 || ^8.0.4
- sentry/sdk: ^3.1
Requires (Dev)
- neos/behat: @dev
README
About
This package provides a Flow integration for the sentry.io PHP SDK. Some basic information about the Flow application is added to the sentry Event by default, but you can easily configure and extend this package to fit your needs.
Installation
composer require netlogix/sentry
Currently the following Flow versions are supported:
^7.3
^8.0
^9.0
Setup
The sentry DSN Client Key has to be configured. Get it from your project settings (SDK Setup -> Client Keys (DSN)).
Netlogix: Sentry: dsn: 'https://fd5c649e6e4d41dd8ca729b15cc5d1c7@o01392.ingest.sentry.io/123456789'
Then simply run ./flow sentry:test
to log an exception to sentry. While this is technically all you have to do,
you might want to adjust the providers - see below.
Configuration
This package allows you to configure which data should be added to the sentry event by changing the providers for each
scope. Currently, the available scopes are environment
, extra
, release
, tags
and user
.
Providers can be sorted using
the PositionalArraySorter
position strings. For the scopes extra
, tags
and user
, all data provided will be merged together. The
scopes environment
and release
only support a single value (you can still configure more than one provider, but
the last one wins).
Netlogix: Sentry: scope: extra: [ ] release: # If you don't need a specific order, you can simply set the provider to true 'Netlogix\Sentry\Scope\Release\PathPattern': true tags: # Numerical order can be used 'Netlogix\Sentry\Scope\Tags\FlowEnvironment': '10' 'Your\Custom\TagProvider': '20' user: 'Your\Custom\UserProvider': 'start 1000' # If you don't want to add the currently authenticated Flow Account to the Event, simply disable the provider 'Netlogix\Sentry\Scope\User\FlowAccount': false
Environments
The sentry SDK will search for the environment variable SENTRY_ENVIRONMENT
and use it's value as the current
environment. This is still the default, however you can configure the Netlogix\Sentry\Scope\Environment\FlowSettings
provider to use a different value:
Netlogix: Sentry: environment: setting: '%env:SENTRY_ENVIRONMENT%'
Release tracking
You can use the Netlogix\Sentry\Scope\Release\PathPattern
ReleaseProvider
to extract your current release from the
app directory. By default, the configured pathPattern
is matched against the FLOW_PATH_ROOT
constant:
Netlogix: Sentry: # Used by Netlogix\Sentry\Scope\Release\PathPattern release: # Path to use for extraction of release pathToMatch: '%FLOW_PATH_ROOT%' # Pattern to extract current release from file path # This pattern is matched against pathToMatch pathPattern: '~/releases/(\d{14})$~'
You can also use the Netlogix\Sentry\Scope\Release\FlowSettings
to set the Release through Flow
Configuration (Netlogix.Sentry.release.setting
, set to %env:SENTRY_RELEASE%
by default).
Custom Providers
For each scope, you can implement your own providers. Each scope requires it's own interface:
- Scope
environment
=>Netlogix\Sentry\Scope\Environment\EnvironmentProvider
- Scope
extra
=>Netlogix\Sentry\Scope\Extra\ExtraProvider
- Scope
release
=>Netlogix\Sentry\Scope\Release\ReleaseProvider
- Scope
tags
=>Netlogix\Sentry\Scope\Tags\TagProvider
- Scope
user
=>Netlogix\Sentry\Scope\User\UserProvider
Then simply add them to the configuration.
If you need access to the thrown exception, you can check Netlogix\Sentry\Scope\ScopeProvider::getCurrentThrowable()
:
<?php namespace Netlogix\Sentry\Scope\Extra; use Neos\Flow\Annotations as Flow; use Neos\Flow\Exception as FlowException; use Netlogix\Sentry\Scope\ScopeProvider; /** * @Flow\Scope("singleton") */ final class ReferenceCodeProvider implements ExtraProvider { private ScopeProvider $scopeProvider; public function __construct(ScopeProvider $scopeProvider) { $this->scopeProvider = $scopeProvider; } public function getExtra(): array { $throwable = $this->scopeProvider->getCurrentThrowable(); if (!$throwable instanceof FlowException) { return []; } return ['referenceCode' => $throwable->getReferenceCode()]; } }
Manually logging exceptions to sentry
If you need to manually send exceptions to sentry (inside a catch
block for example), you can use the
Netlogix\Sentry\ThrowableStorage\SentryStorage
:
<?php use Neos\Flow\Annotations as Flow; use Netlogix\Sentry\ThrowableStorage\SentryStorage; class LoggingManually { /** * @Flow\Inject * @var SentryStorage */ protected $sentryStorage; public function log(): void { $exception = new \RuntimeException('foo', 1612114936); $this->sentryStorage->logThrowable($exception, ['some' => ['additional', 'data']]); } }
Ignoring exceptions
If you need to skip sending a specific exception to sentry, you can use Flow's renderingGroups
. Simply create one that
matches your exception and set logException
to false
:
Neos: Flow: error: exceptionHandler: renderingGroups: ignoredExceptions: matchingStatusCodes: [ 418 ] matchingExceptionClassNames: [ 'Your\Ignored\Exception' ] # It is also possible to match against \Throwable::getCode(). Please note that this is not a Flow feature. # Check \Netlogix\Sentry\ExceptionHandler\ExceptionRenderingOptionsResolver::resolveRenderingGroup() for more info # matchingExceptionCodes: [1638880375] options: logException: false
Please note that this also disables logging of this exception to Data/Logs/Exceptions
.
Encrypt POST payload
By default, the array of POST payload data is transported to the sentry server "as is".
When encryption is enabled and a valid rsa key fingerprint is set, the POST payload is stripped and replaced by an RSA encrypted string.
Netlogix: Sentry: privacy: encryptPostBody: true rsaKeyFingerprint: '6ff568ae0f9b44b69627e275accf163a'
POST data without encryption usually looks like this in sentry:
{ "--some-form": { "__currentPage": 1, "__state": "TmV0bG9naXguU2VudHJ5IHN0YXRlIGRhdGE=", "__trustedProperties": "[Filtered]", "firstName": "John", "lastName": "Doe", "birthday": "2021-01-01", "email": "john.doe@netlogix.de", "message": "Lorem ipsum dolor sit amet" } }
With encryption enabled it looks like this:
{ "__ENCRYPTED__DATA__": { "encryptedData": "TG9yZW0gaXBzdW0gZG9sb3Igc2l0IGFtZXQsIGNvbnNlY3RldHVyIGFkaXBpc2ljaSBlbGl0LCBzZWQgZWl1c21vZCB0ZW1wb3IgaW5jaWR1bnQgdXQgbGFib3JlIGV0IGRvbG9yZSBtYWduYSBhbGlxdWEuIFV0IGVuaW0gYWQgbWluaW0gdmVuaWFtLCBxdWlzIG5vc3RydWQgZXhlcmNpdGF0aW9u", "envelopeKey": "ZGVzZXJ1bnQgbW9sbGl0IGFuaW0gaWQgZXN0IGxhYm9ydW0=", "initializationVector": "QmxpbmR0ZXh0" } }
There will be an additional sentry field "Encrypted POST Data" which contains a backlink to encrypt and show the original data.
In order for this to work, there must be an authentication provider in place that handels the Neos.Sentry controller.
If this package is used in conjunction with the neos/neos CMS, the neos backend authentication provider can be tasked with this job. See the code snippet below.
If this package is used without neos/neos, a custom privilege for policy Netlogix.Sentry:Backend.EncryptedPayload
has to be configured.
Neos: Flow: security: authentication: providers: 'Neos.Neos:Backend': requestPatterns: 'Netlogix.Sentry:ShowEncryptedPayload': pattern: ControllerObjectName patternOptions: controllerObjectNamePattern: 'Netlogix\Sentry\Controller\.*'
Scrubbing variables or adding more details
Allow in php.ini config
Context variables depend on zend not removing exception arguments. Make sure zend is not configured to remove arguments from exceptions to make use of fine-grained variable scrubbing and context-aware variable details.
[php]
zend.exception_ignore_args=0
Scrub all frame arguments
Ignoring exception arguments via zend setting would just remove every exception argument. Scrubbing can be re-added manually in userspace code by config settings. This might be helpful for either different environments or deployment targets.
Netlogix: Sentry: variableScrubbing: scrubbing: true
Prevent specific arguments from being scrubbed
In contrast to zend.exception_ignore_args, having them scrubbed manually allows for keeping very specific variables by naming them individually.
Netlogix: Sentry: variableScrubbing: # Only works when manual scrubbing is enabled. Otherwise, all data is # kept anyway. scrubbing: true # Keep certain frame variables even if scrubbing is enabled # # In contrast to using an individual RepresentationSerializer, this is context # aware. Not every string should be displayed, but some might. # # Complex arguments might be obfuscated when being passed through the default # Sentry RepresentationSerializerInterface::representationSerialize() and not very # useful. Use our VariablesFromStackProvider and the "contextDetails" setting for # more detailed logging. keepFromScrubbing: 'Neos\ContentRepository\Search\Indexer\NodeIndexingManager::indexNode()': # Class names cover "_Original" as well, but neither implementing # interfaces nor extending classes. This only matches by exact string # comparison. className: 'Neos\ContentRepository\Search\Indexer\NodeIndexingManager' methodName: 'indexNode' # Arguments can only be argument names, not argument paths. Preventing # certain arguments form being scrubbed happens after the Sentry # RepresentationSerializer converted them, so no actual object # information is available for detailed matching. arguments: - 'node'
Add specific object information of frame variables to the sentry event context
Data that is not ony "a representation of" a method argument but somewhere within any given data structure can be extracted and adde to the sentry event.
This covers way more details about any given event frame, as long as the information is accessible by either public properties or a getter method.
Netlogix: Sentry: variableScrubbing: # In contrast to scrubbing and keeping specific arguments from being # scrubbed, this context vars are added "in full" to the event context. contextDetails: 'Neos\ContentRepository\Search\Indexer\NodeIndexingManager::indexNode()': className: 'Neos\ContentRepository\Search\Indexer\NodeIndexingManager' methodName: 'indexNode' arguments: # It's necessary to specify exactly which property paths should be # added to prevent loosing information. Data not being specified is # just dropped. - 'node.path' - 'node.identifier' - 'node.name'