We'd love your help!
General contributing guidelines are described in Contributing Guidelines.
Jaeger is Apache 2.0 licensed and accepts contributions via GitHub pull requests. This document outlines some of the conventions on development workflow, commit message formatting, contact points and other resources to make it easier to get your contribution accepted.
We gratefully welcome improvements to documentation as well as to code.
- Install Go and setup GOPATH and add $GOPATH/bin in PATH
- Install Yarn for running local build with the UI
This library uses Go modules to manage dependencies.
git clone [email protected]:jaegertracing/jaeger.git jaeger
cd jaeger
Then install dependencies and run the tests:
# Adds the jaeger-ui submodule
git submodule update --init --recursive
# Installs required tools
make install-tools
# Runs all unit tests
make test
$ make run-all-in-one
The jaeger-ui
submodule, which was added from the Pre-requisites step above, contains
the source code for the UI assets (requires Node.js 6+).
The assets must be compiled first with make build-ui
, which runs Node.js build and then
packages the assets into a Go file that is .gitignore
-ed.
The packaged assets can be enabled by providing a build tag ui
, for example:
$ go run -tags ui ./cmd/all-in-one/main.go
make run-all-in-one
essentially runs Jaeger all-in-one by combining both of the above
steps into a single make
command.
These are general guidelines on how to organize source code in this repository.
github.com/jaegertracing/jaeger
cmd/ - All binaries go here
agent/
app/ - The actual code for the binary
main.go
collector/
app/ - The actual code for the binary
main.go
crossdock/ - Cross-repo integration test configuration
examples/
hotrod/ - Demo application that uses OpenTracing API
idl/ - (submodule) https://github.com/jaegertracing/jaeger-idl
jaeger-ui/ - (submodule) https://github.com/jaegertracing/jaeger-ui
model/ - Where models are kept, e.g. Process, Span, Trace
pkg/ - (See Note 1)
plugin/ - Swappable implementations of various components
storage/
cassandra/ - Cassandra implementations of storage APIs
. - Shared Cassandra stuff
spanstore/ - SpanReader / SpanWriter implementations
dependencystore/
elasticsearch/ - ES implementations of storage APIs
scripts/ - Miscellaneous project scripts, e.g. github action and license update script
storage/
spanstore/ - SpanReader / SpanWriter interfaces
dependencystore/
thrift-gen/ - Generated Thrift types
agent/
jaeger/
sampling/
zipkincore/
go.mod - Go module file to track dependencies
- Note 1:
pkg
is a collection of utility packages used by the Jaeger components without being specific to its internals. Utility packages are kept separate from the Jaeger core codebase to keep it as small and concise as possible. If some utilities grow larger and their APIs stabilize, they may be moved to their own repository, to facilitate re-use by other projects.
This projects follows the following pattern for grouping imports in Go files:
- imports from standard library
- imports from other projects
- imports from
jaeger
project
For example:
import (
"fmt"
"github.com/uber/jaeger-lib/metrics"
"go.uber.org/zap"
"github.com/jaegertracing/jaeger/cmd/agent/app"
"github.com/jaegertracing/jaeger/cmd/collector/app/builder"
)
We strive to maintain as high code coverage as possible. The current repository limit is set at 95%, with some exclusions discussed below.
Since go test
command does not generate
code coverage information for packages that have no test files, we have a build step (make nocover
)
that breaks the build when such packages are discovered, with the following error:
error: at least one *_test.go file must be in all directories with go files
so that they are counted for code coverage.
If no tests are possible for a package (e.g. it only defines types), create empty_test.go
As the message says, all packages are required to have at least one *_test.go
file.
There are conditions that cannot be tested without external dependencies, such as a function that
creates a gocql.Session
, because it requires an active connection to Cassandra database. It is
recommended to isolate such functions in a separate package with bare minimum of code and add a
file .nocover
to exclude the package from coverage calculations. The file should contain
a comment explaining why it is there, for example:
$ cat ./pkg/cassandra/config/.nocover
requires connection to Cassandra
Before merging a PR make sure:
- the title is descriptive and follows a good commit message
- pull request is assigned to the current release milestone
- add
changelog:*
and other labels
Merge the PR by using "Squash and merge" option on Github. Avoid creating merge commits. After the merge make sure referenced issues were closed.
- If a flag is deprecated in release N, it can be removed in release N+2 or three months later, whichever is later.
- When adding a (deprecated) prefix to the flags, indicate via a deprecation message that the flag could be removed in the future. For example:
(deprecated, will be removed after 2020-03-15 or in release v1.19.0, whichever is later)
- At the top of the file where the flag name is defined, add a constant and a comment, e.g.
// TODO deprecated flag to be removed healthCheckHTTPPortWarning = "(deprecated, will be removed after 2020-03-15 or in release v1.19.0, whichever is later)"
- Use that constant as the prefix to the help text, e.g.
flagSet.Int(healthCheckHTTPPort, 0, healthCheckHTTPPortWarning+" see --"+adminHTTPHostPort)
- When parsing a deprecated flag into config, log a warning with the same deprecation message
- Take care of deprecated flags in
initFromViper
functions, do not pass them to business functions.
- Ensure all references to the flag's variables have been removed in code.
- Ensure a "Breaking Changes" entry is added in the CHANGELOG indicating which CLI flag is being removed and which CLI flag should be used in favor of this removed flag.
For example:
* Remove deprecated flags `--old-flag`, please use `--new-flag` ([#1234](<pull-request URL>), [@myusername](https://github.com/myusername))