Update sentry gems
What does this MR do and why?
Updates Sentry gems to the latest version.
The change in 5.2.0 would not fix the issue of oversaturating Redis, but it seems that maintainers work on improving performance tracing in general. Having a fresh gem would be great for future work in that direction
🔎 Click to expand - Changelog diff
## 5.8.0
Features
-
Allow tags to be passed via the context hash when reporting errors using ActiveSupport::ErrorReporter and Sentry::Rails::ErrorSubscriber in
sentry-rails
#1932 -
Pass a
cached: true
tag for SQL query spans that utilized the ActiveRecord QueryCache when using ActiveRecordSubscriber insentry-rails
#1968 -
Add
Sentry.add_global_event_processor
API #1976Users can now configure global event processors without configuring scope as well.
Sentry.add_global_event_processor do |event, hint| event.tags = { foo: 42 } event end
-
Add global event processor in OpenTelemetry
SpanProcessor
to link errors with transactions #1983 -
Fix some inconsistencies in setting name/op/status in OpenTelemetry
SpanProcessor
#1987 -
Add
config.before_send_transaction
hook #1989Users can now configure a
before_send_transaction
callback that runs similar tobefore_send
but for transaction events.config.before_send_transaction = lambda do |event, hint| # skip unimportant transactions or strip sensitive data if event.transaction == "/healthcheck/route" nil else event end end
-
Support
Sentry::Transaction#set_measurement
#1838Usage:
transaction = Sentry.get_current_scope.get_transaction transaction.set_measurement("metrics.foo", 0.5, "millisecond")
Bug Fixes
- Support redis-rb 5.0+ #1963
- Fixes #1932
- Skip private _config context in Sidekiq 7+ #1967
- Fixes #1956
- Return value from
perform_action
in ActionCable::Channel instances when initialized #1966 -
Span#with_child_span
should finish the span even with exception raised #1982 - Fix sentry-rails' controller span nesting #1973
- Fixes #1899
- Do not report exceptions when a Rails runner exits with
exit 0
#1988 - Ignore redis key if not UTF8 #1997
- Fixes #1992
Miscellaneous
- Deprecate
capture_exception_frame_locals
in favor ofinclude_local_variables
#1993
5.7.0
Features
- Expose
span_id
inSpan
constructor #1945 - Expose
end_timestamp
inSpan#finish
andTransaction#finish
#1946 - Add
Transaction#set_context
api #1947 - Add OpenTelemetry support with new
sentry-opentelemetry
gem-
Add
config.instrumenter
to switch between:sentry
and:otel
instrumentation #1944The new
sentry-opentelemetry
gem adds support to automatically integrate OpenTelemetry performance tracing with Sentry. Give it a try and let us know if you have any feedback or problems with using it.
-
5.6.0
Features
-
Allow users to configure their asset-skipping pattern #1915
Users can now configure their own pattern to skip asset requests' transactions
Sentry.init do |config| config.rails.assets_regexp = /my_regexp/ end
-
Use
Sentry.with_child_span
in redis and net/http instead ofspan.start_child
#1920- This might change the nesting of some spans and make it more accurate
- Followup fix to set the sentry-trace header in the correct place #1922
-
Use
Exception#detailed_message
when generating exception message if applicable #1924 -
Make
sentry-sidekiq
compatible with Sidekiq 7 #1930
Bug Fixes
-
Sentry::BackgroundWorker
will releaseActiveRecord
connection pool only when theActiveRecord
connection is established -
Remove bad encoding arguments in redis span descriptions #1914
- Fixes #1911
-
Add missing
initialized?
checks tosentry-rails
#1919- Fixes #1885
-
Update Tracing Span's op names #1923
Currently, Ruby integrations' Span op names aren't aligned with the core specification's convention, so we decided to update them altogether in this PR. If you rely on Span op names for fine-grained event filtering, this may affect the data your app sends to Sentry. Also make sure to update your
traces_sampler
if you rely on theop
for filtering some requests.
Refactoring
- Make transaction a required argument of Span #1921
5.5.0
Features
-
Support rack 3 #1884
- We no longer need the
HTTP_VERSION
check for ignoring the header
- We no longer need the
-
Add Dynamic Sampling support The SDK now supports Sentry's Dynamic Sampling product.
Note that this is not supported for users still using the
config.async
option.- Parse incoming W3C Baggage Headers and propagate them to continue traces #1869
- in all outgoing requests in our net/http patch
- in Sentry transactions as Dynamic Sampling Context
- Create new Baggage entries as Head SDK (originator of trace) #1898
- Add Transaction source annotations to classify low quality (high cardinality) transaction names #1902
- Parse incoming W3C Baggage Headers and propagate them to continue traces #1869
Bug Fixes
- Memoize session.aggregation_key #1892
- Fixes #1891
- Execute
with_scope
's block even when SDK is not initialized #1897- Fixes #1896
- Make sure test helper clears the current scope before/after a test #1900
5.4.2
Bug Fixes
5.4.1
Bug Fixes
- Fix missing
spec.files
insentry-ruby.gemspec
- Fixes #1856
5.4.0
Features
-
Expose
:values
inExceptionInterface
, so that it can be accessed inbefore_send
underevent.exception.values
#1843 -
Add top level
Sentry.close
API #1844- Cleans up SDK state and sets it to uninitialized
- No-ops all SDK APIs and also disables the transport layer, so nothing will be sent to Sentry after closing the SDK
-
Handle exception with large stacktrace without dropping entire item #1807
-
Capture Rails runner's exceptions before exiting #1820
-
Add
Sentry.with_exception_captured
helper #1814Usage:
Sentry.with_exception_captured do 1/1 #=> 1 will be returned end Sentry.with_exception_captured do 1/0 #=> ZeroDivisionError will be reported and re-raised end
-
Prepare for Rails 7.1's error reporter API change #1834
-
Set
sentry.error_event_id
in request env if the middleware captures errors #1849If the SDK's Rack middleware captures an error, the reported event's id will be stored in the request env. For example:
env["sentry.error_event_id"] #=> "507bd4c1a07e4355bb70bcd7afe8ab17"
Users can display this information on the error page via a middleware as proposed in #1846
Bug Fixes
- Respect
report_rescued_exceptions
config #1847- Fixes #1840
- Rescue event's to JSON conversion error #1853
- Rescue
ThreadError
inSessionFlusher
and stop creating threads if flusher is killed #1851- Fixes #1848
Refactoring
- Move envelope item processing/trimming logic to the Item class #1824
- Replace sentry-ruby-core with sentry-ruby as integration dependency #1825
Test Helpers
The SDK now provides a set of test helpers to help users setup and teardown Sentry related tests.
To get started:
require "sentry/test_helper"
# in minitest
class MyTest < Minitest::Test
include Sentry::TestHelper
# ...
end
# in RSpec
RSpec.configure do |config|
config.include Sentry::TestHelper
# ...
end
It's still an early attempt so please give us feedback in #1680.
5.3.1
Bug Fixes
- Don't require a DB connection, but release one if it is acquired #1812
- Fixes #1808
-
Sentry.with_child_span
should check SDK's initialization state #1819- Fixes #1818
Miscellaneous
- Warn users about
config.async
's deprecation #1803
5.3.0
Features
- Add
Sentry.with_child_span
for easier span recording #1783
operation_result = Sentry.with_child_span(op: "my op") do |child_span|
my_operation
end
# the "my op" span will be attached to the result of Sentry.get_current_scope.get_span
# which could be either the top-level transaction, or a span set by the user or other integrations
Bug Fixes
- Set
last_event_id
only for error events #1767- Fixes #1766
- Add
config.rails.register_error_subscriber
to control error reporter integration #1771 - Check if ActiveRecord connection exists before calling AR connection pool #1769
- Fixes #1745
- Fix
sentry-rails
's tracing spans not nesting issue - #1784- Fixes #1723
- Update
config.transport.proxy
to allow String and URI values as previously supported bysentry-ruby
versions <= 4.8 using Faraday- Fixes #1782
- Register SentryContextClientMiddleware on sidekiq workers #1774
- Add request env to sampling context when using
sentry-rails
#1792- Fixes #1791
- Fix net-http tracing's span nesting issue #1796
Refactoring
- Correct inaccurate event model relationships #1777
Miscellaneous
- Log message when shutting down/killing SDK managed components #1779
5.2.1
Bug Fixes
- Also check stringified breadcrumbs key when reducing payload size #1758
- Fixes #1757
- Ignore internal Sidekiq::JobRetry::Skip exception #1763
- Fixes #1731
Miscellaneous
- Warn user if any integration is required after SDK initialization #1759
5.2.0
Features
-
Log Redis command arguments when sending PII is enabled #1726
-
Add request env to sampling context #1749
Example
Sentry.init do |config| config.traces_sampler = lambda do |sampling_context| env = sampling_context[:env] if env["REQUEST_METHOD"] == "GET" 0.01 else 0.1 end end end
-
Check envelope size before sending it #1747
The SDK will now check if the envelope's event items are oversized before sending the envelope. It goes like this:
- If an event is oversized (200kb), the SDK will remove its breadcrumbs (which in our experience is the most common cause).
- If the event size now falls within the limit, it'll be sent.
- Otherwise, the event will be thrown away. The SDK will also log a debug message about the event's attributes size (in bytes) breakdown. For example,
{event_id: 34, level: 7, timestamp: 22, environment: 13, server_name: 14, modules: 935, message: 5, user: 2, tags: 2, contexts: 820791, extra: 2, fingerprint: 2, platform: 6, sdk: 40, threads: 51690}
This will help users report size-related issues in the future.
-
Automatic session tracking #1715
Example:
The SDK now supports automatic session tracking / release health by default in Rack based applications. Aggregate statistics on successful / errored requests are collected and sent to the server every minute. To use this feature, make sure the SDK can detect your app's release. Or you have set it with:
Sentry.init do |config| config.release = 'release-foo-v1' end
To disable this feature, set
config.auto_session_tracking
tofalse
.
Bug Fixes
MR acceptance checklist
This checklist encourages us to confirm any changes have been analyzed to reduce risks in quality, performance, reliability, security, and maintainability.
-
I have evaluated the MR acceptance checklist for this MR.