Switch branches/tags
Find file Copy path
283 lines (207 sloc) 11.2 KB

Contributing to InfluxDB

Bug reports

Before you file an issue, please search existing issues in case it has already been filed, or perhaps even fixed. If you file an issue, please include the following.

  • Full details of your operating system (or distribution) e.g. 64-bit Ubuntu 14.04.
  • The version of InfluxDB you are running
  • Whether you installed it using a pre-built package, or built it from source.
  • A small test case, if applicable, that demonstrates the issues.

Remember the golden rule of bug reports: The easier you make it for us to reproduce the problem, the faster it will get fixed. If you have never written a bug report before, or if you want to brush up on your bug reporting skills, we recommend reading Simon Tatham's essay "How to Report Bugs Effectively."

Test cases should be in the form of curl commands. For example:

# create database
curl -X POST http://localhost:8086/query --data-urlencode "q=CREATE DATABASE mydb"

# create retention policy
curl -X POST http://localhost:8086/query --data-urlencode "q=CREATE RETENTION POLICY myrp ON mydb DURATION 365d REPLICATION 1 DEFAULT"

# write data
curl -X POST http://localhost:8086/write?db=mydb --data-binary "cpu,region=useast,host=server_1,service=redis value=61"

# Delete a Measurement
curl -X POST http://localhost:8086/query  --data-urlencode 'db=mydb' --data-urlencode 'q=DROP MEASUREMENT cpu'

# Query the Measurement
# Bug: expected it to return no data, but data comes back.
curl -X POST http://localhost:8086/query  --data-urlencode 'db=mydb' --data-urlencode 'q=SELECT * from cpu'

If you don't include a clear test case like this, your issue may not be investigated, and may even be closed. If writing the data is too difficult, please zip up your data directory and include a link to it in your bug report.

Please note that issues are not the place to file general questions such as "how do I use collectd with InfluxDB?" Questions of this nature should be sent to the InfluxData Community, not filed as issues. Issues like this will be closed.

Feature requests

We really like to receive feature requests, as it helps us prioritize our work. Please be clear about your requirements, as incomplete feature requests may simply be closed if we don't understand what you would like to see added to InfluxDB.

Contributing to the source code

InfluxDB follows standard Go project structure. This means that all your Go development are done in $GOPATH/src. GOPATH can be any directory under which InfluxDB and all its dependencies will be cloned. For full details on the project structure, follow along below.

You should also read our coding guide, to understand better how to write code for InfluxDB.

Submitting a pull request

To submit a pull request you should fork the InfluxDB repository, and make your change on a feature branch of your fork. Then generate a pull request from your branch against master of the InfluxDB repository. Include in your pull request details of your change -- the why and the how -- as well as the testing your performed. Also, be sure to run the test suite with your change in place. Changes that cause tests to fail cannot be merged.

There will usually be some back and forth as we finalize the change, but once that completes it may be merged.

To assist in review for the PR, please add the following to your pull request comment:

- [ ] updated
- [ ] Rebased/mergable
- [ ] Tests pass
- [ ] Sign [CLA]( (if not already signed)

Signing the CLA

If you are going to be contributing back to InfluxDB please take a second to sign our CLA, which can be found on our website.

Installing Go

InfluxDB requires Go 1.11.

At InfluxDB we find gvm, a Go version manager, useful for installing Go. For instructions on how to install it see the gvm page on github.

After installing gvm you can install and set the default go version by running the following:

gvm install go1.11
gvm use go1.11 --default

Installing Dep

InfluxDB uses dep to manage dependencies. Install it by running the following:

go get

Revision Control Systems

Go has the ability to import remote packages via revision control systems with the go get command. To ensure that you can retrieve any remote package, be sure to install the following rcs software to your system. Currently the project only depends on git and mercurial.

Getting the source

Setup the project structure and fetch the repo like so:

    mkdir $HOME/gocodez
    export GOPATH=$HOME/gocodez
    go get

You can add the line export GOPATH=$HOME/gocodez to your bash/zsh file to be set for every shell instead of having to manually run it everytime.

Cloning a fork

If you wish to work with fork of InfluxDB, your own fork for example, you must still follow the directory structure above. But instead of cloning the main repo, instead clone your fork. Follow the steps below to work with a fork:

    export GOPATH=$HOME/gocodez
    mkdir -p $GOPATH/src/
    cd $GOPATH/src/
    git clone<username>/influxdb

Retaining the directory structure $GOPATH/src/ is necessary so that Go imports work correctly.

Build and Test

Make sure you have Go installed and the project structure as shown above. To then get the dependencies for the project, execute the following commands:

cd $GOPATH/src/
dep ensure

To then build and install the binaries, run the following command.

go clean ./...
go install ./...

The binaries will be located in $GOPATH/bin. Please note that the InfluxDB binary is named influxd, not influxdb.

To set the version and commit flags during the build pass the following to the install command:

-ldflags="-X main.version=$VERSION -X main.branch=$BRANCH -X main.commit=$COMMIT"

where $VERSION is the version, $BRANCH is the branch, and $COMMIT is the git commit hash.

If you want to build packages, see usage information:

python --help

# Or to build a package for your current system
python --package

To run the tests, execute the following command:

cd $GOPATH/src/
go test -v ./...

# run tests that match some pattern
go test -run=TestDatabase . -v

# run tests and show coverage
go test -coverprofile /tmp/cover . && go tool cover -html /tmp/cover

To install go cover, run the following command:

go get

Generated Google Protobuf code

Most changes to the source do not require that the generated protocol buffer code be changed. But if you need to modify the protocol buffer code, you'll first need to install the protocol buffers toolchain.

First install the protocol buffer compiler 2.6.1 or later for your OS:

Then install the go plugins:

go get
go get
go get

Finally run, go generate after updating any *.proto file:

go generate ./...


If generating the protobuf code is failing for you, check each of the following:

  • Ensure the protobuf library can be found. Make sure that LD_LIBRARY_PATH includes the directory in which the library has been installed.
  • Ensure the command protoc-gen-gogo, found in GOPATH/bin, is on your path. This can be done by adding GOPATH/bin to PATH.

Generated Go Templates

The query engine requires optimized data structures for each data type so instead of writing each implementation several times we use templates. Do not change code that ends in a .gen.go extension! Instead you must edit the .gen.go.tmpl file that was used to generate it.

Once you've edited the template file, you'll need the tmpl utility to generate the code:

$ go get

Then you can regenerate all templates in the project:

$ go generate ./...

Pre-commit checks

We have a pre-commit hook to make sure code is formatted properly and vetted before you commit any changes. We strongly recommend using the pre-commit hook to guard against accidentally committing unformatted code. To use the pre-commit hook, run the following:

    cd $GOPATH/src/
    cp .hooks/pre-commit .git/hooks/

In case the commit is rejected because it's not formatted you can run the following to format the code:

go fmt ./...
go vet ./...

To install go vet, run the following command:

go get

NOTE: If you have not installed mercurial, the above command will fail. See Revision Control Systems above.

For more information on go vet, read the GoDoc.


When troubleshooting problems with CPU or memory the Go toolchain can be helpful. You can start InfluxDB with CPU and memory profiling turned on. For example:

# start influx with profiling
./influxd -cpuprofile -memprof
# run queries, writes, whatever you're testing
# Quit out of influxd and will then be written.
# open up pprof to examine the profiling data.
go tool pprof ./influxd
# once inside run "web", opens up browser with the CPU graph
# can also run "web <function name>" to zoom in. Or "list <function name>" to see specific lines

Note that when you pass the binary to go tool pprof you must specify the path to the binary.

If you are profiling benchmarks built with the testing package, you may wish to use the package to limit the code being profiled:

func BenchmarkSomething(b *testing.B) {
  // do something intensive like fill database with data...
  defer profile.Start(profile.ProfilePath("/tmp"), profile.MemProfile).Stop()
  // do something that you want to profile...

Continuous Integration testing

InfluxDB uses CircleCI for continuous integration testing. CircleCI executes, so you may do the same on your local development environment before creating a pull request.

The script executes a test suite with 5 variants (standard 64 bit, 64 bit with race detection, 32 bit, TSI, go version 1.11), each executes with a different arg, 0 through 4. Unless you know differently, ./ 0 is probably all you need.