Starting up a BuildGrid server requires only installation and a configuration file.
WARNING: The following setup is highly insecure and intended for proof-of-concept testing ONLY! A malicious user can have full unsandboxed access to the entire host system. A secure setup must be configured with authentication.
If you’d like to get started, a sample configuration that uses your filesystem as the CAS (Content Addressable Storage) and a SQLite-based scheduler is provided below. Copy the following text into a file called config.yml and change /path/to/cas/ to a suitable storage path.
By default the scheduler’s data store is written to ./example.db. You can also change this to a suitable storage path.
server: - !channel port: 50051 insecure-mode: true description: > BuildGrid's default configuration: - Unauthenticated plain HTTP at :50051 - Single instance: [unnamed] - In-memory data, max. 2Gio - DataStore: sqlite:///./example.db - Hosted services: - ActionCache - Execute - ContentAddressableStorage - ByteStream authorization: method: none monitoring: enabled: false instances: - name: '' description: | The unique '' instance. storages: - !disk-storage &cas-storage path: /path/to/cas/ schedulers: - !sql-scheduler &state-database storage: *cas-storage connection-string: sqlite:///./example.db automigrate: yes connection-timeout: 15 poll-interval: 0.5 services: - !action-cache &build-cache storage: *cas-storage max-cached-refs: 256 cache-failed-actions: true allow-updates: true - !execution storage: *cas-storage action-cache: *build-cache scheduler: *state-database - !cas storage: *cas-storage - !bytestream storage: *cas-storage
To start buildgrid with this configuration, simply run:
bgd server start --verbose /path/to/config.yml
See the Understanding the configuration file section to learn more about this file. For now, we will continue setting up BuildGrid for work.
Setting up a bot¶
Now we will need a worker. The recommended worker to use with BuildGrid is buildbox-worker. This worker works best when used alongside a local CAS cache called buildbox-casd. First, build these tools following the instructions in their READMEs.
Then, start the CAS cache.
buildbox-casd --cas-remote=http://localhost:50051 --bind=127.0.0.1:50011 ~/casd &
Once CASD is running we can start the worker itself, pointing it to CASD for CAS requests.
buildbox-worker --buildbox-run=buildbox-run-hosttools --bots-remote=http://localhost:50051 \ --cas-remote=http://127.0.0.1:50011 --request-timeout=30 my_bot
We should be able to see this worker connecting as log messages for
UpdateBotSession requests in the server logs.
Whilst this approach has less moving parts, it will make your build slower due to
needing to freshly fetch the input root for every Action rather than keeping a local
cache. Production deployments should use
buildbox-worker supports running without
buildbox-casd by pointing it to the remote CAS
rather than the local CASD, although this isn’t recommended due to the additional network load
it will lead to. When running in this configuration, its important to tell the runner command
to not use the LocalCAS protocol to stage the input root.
buildbox-worker --buildbox-run=buildbox-run-hosttools --bots-remote=http://localhost:50051 \ --cas-remote=http://localhost:50051 --request-timeout=30 --runner-arg=--disable-localcas my_bot
This worker is only recommended for use in quick proof-of-concept tests and development purposes, to verify your Execution service is working and can hand off Actions to workers without needing to build the buildbox tools. Deployments executing real Actions are recommended to use buildbox-worker and buildbox-casd.
BuildGrid comes with a test worker called “bgd bot” that can be used to execute jobs on the same system as the service itself with no sandboxing. In a new terminal, run the following command:
bgd bot --verbose --remote http://localhost:50051 --remote-cas http://localhost:50051 host-tools
In this case, our Execution service and CAS service are located at the same endpoint, so we can pass the same address to both flags.
Sending work to BuildGrid¶
We have a working BuildGrid setup! Now you can send work to it.
You can use your own remote execution compatible client, but BuildGrid also comes with a
command-line tool for execute simple commands called
We’re going to send a simple job that just cats a file. All Execute requests specify an
Action to execute, and each Action has an “input root,” which is a directory that the job
is performed inside.
bgd execute requires this input root as a command-line parameter,
so let’s make one.
In another terminal, create a directory in your home directory or somewhere else that is convenient, then add a file to it:
mkdir ~/my_input_root echo "I'm in the input root!" >> ~/my_input_root/input_root.txt
Now, let’s send the action to our running BuildGrid instance and bot with “bgd execute.”
bgd execute --remote http://localhost:50051 --remote-cas http://localhost:50051 command ~/my_input_root /bin/cat input_root.txt
In the above request, notice that the input root (
~/my_input_root) is specified first, before
the rest of the command. The file is specified relative to the input root.
If all goes well, the request should have been sent to BuildGrid, which will have
farmed it out to the bot. The bot will have done the work and sent it back to BuildGrid,
and bgd execute should display the response metadata. Look for the
which will contain the text of the file we catted.
stdout_raw field is empty, the output might be stored in CAS and need to be fetched.
The digest to fetch is in the
stdout_digest field. We can fetch it using the
bgd cas --remote http://localhost:50051 download-file <stdout_digest> action-stdout.log
Understanding the configuration file¶
Looking at the config file provides insight into BuildGrid’s structure. While a full breakdown of the configuration options can be found on the configuration page, the “instances” section deserves special mention.
BuildGrid is not just a single service, but rather a collection of services that work together to facilitate remote execution. The REAPI has the notion of “instances”, which effectively act as namespaces for requests since requests must contain the instance name as a field. Each instance can have one or more services attached to it.
The services that can be attached to an instance are as follows:
The Execution Service is BuildGrid’s primary service. Your clients’ Execute() requests are sent here for processing. Note that it checks the ActionCache to determine whether it can just return a cached result. The data-store represents the backend for the scheduling system. In this case, we’re using a sql-backed scheduler.
This is the interface into the Content-Addressable Storage. Note that the other services talk to the storage backend directly rather than make requests to this service.
This is the interface into the ByteStream service. This is used for blobs too large to be handled by the batch gRPC methods of the CAS service.
The ActionCache service handles requests to the ActionCache. It needs to talk to the backend storage to retrieve the ActionResult blobs. In this configuration, we have set the maximum number of cacheable actions to 256, and we allow failed builds to be cached and make the ActionCache writeable.
This is a BuildStream-specific service built on top of the CAS service. It is used to associate names with blobs in CAS. For more information, see the BuildStream proto.
Now that we have a working BuildGrid, it is time to run some real Actions. Usage of BuildGrid with various clients is documented in the Using section, along with information on how to use the sandboxing features of BuildBox and some of the more advanced features of BuildGrid such as Indexed CAS and CAS cleanup.