Version v1.39 of the documentation is no longer actively maintained. The site that you are currently viewing is an archived snapshot. For up-to-date documentation, see the latest version.
Skaffold API beta
When running skaffold dev
or skaffold debug
,
Skaffold starts a server that exposes an API over the lifetime of the Skaffold process.
Besides the CLI, this API is the primary way tools like IDEs integrate with Skaffold for retrieving information about the
pipeline and for controlling the phases in the pipeline.
To retrieve information about the Skaffold pipeline, the Skaffold API provides two main functionalities:
-
A streaming event log created from the different phases in a pipeline run, and
-
A snapshot of the overall state of the pipeline at any given time during the run.
To control the individual phases of the Skaffold, the Skaffold API provides fine-grained control over the individual phases of the pipeline (build, deploy, and sync).
Connecting to the Skaffold API
The Skaffold API is gRPC
based, and it is also exposed via the gRPC gateway as a JSON over HTTP service.
The API can be enabled via setting the --rpc-port
or --rpc-http-port
flags (or both)
depending on whether you want to enable the gRPC API or the HTTP REST API, respectively.
Note
The--enable-rpc
flag is now deprecated in favor of --rpc-port
and --rpc-http-port
flags.
For reference, we generate the server’s gRPC service definitions and message protos as well as the Swagger based HTTP API Spec.
gRPC Server
The gRPC API can be started by specifying the --rpc-port
flag. If the specified port is not available, Skaffold will
exit with failure.
HTTP server
The HTTP REST API can be started by specifying the --rpc-http-port
flag. If the specified port is not available,
Skaffold will exit with failure.
Starting the HTTP REST API will also start the gRPC API as it proxies the requests to the gRPC API. By default, Skaffold chooses a random available port for gRPC, but it can be customized (see below).
Creating a gRPC Client
To connect to the gRPC
server at the specified port, create a client using the following code snippet.
Note
The skaffold gRPC server is not compatible with HTTPS, so connections need to be marked as insecure withgrpc.WithInsecure()
import (
"log"
pb "github.com/GoogleContainerTools/skaffold/proto/v1"
"google.golang.org/grpc"
)
func main(){
conn, err := grpc.Dial("localhost:50051", grpc.WithInsecure())
if err != nil {
log.Fatalf("fail to dial: %v", err)
}
defer conn.Close()
client := pb.NewSkaffoldServiceClient(conn)
}
API Structure
Skaffold’s API exposes the three main endpoints:
- Event API - continuous stream of lifecycle events
- State API - retrieve the current state
- Control API - control build/deploy/sync
Event API
Skaffold provides a continuous development mode, skaffold dev
, which rebuilds and redeploys
your application on changes. In a single development loop, one or more container images
may be built and deployed.
Skaffold exposes events for clients to be notified when phases within a development loop start, succeed, or fail. Tools that integrate with Skaffold can use these events to kick off parts of a development workflow depending on them.
Example scenarios:
- port-forwarding events are used by Cloud Code to automatically attach debuggers to running containers.
- using an event indicating a frontend service has been deployed and port-forwarded successfully to kick off a suite of Selenium tests against the newly deployed service.
Event API Contract
protocol | endpoint | encoding |
---|---|---|
HTTP | http://localhost:{HTTP_RPC_PORT}/v1/events |
newline separated JSON using chunk transfer encoding over HTTP |
gRPC | client.Events(ctx) method on the SkaffoldService |
protobuf 3 over HTTP |
Examples
Using curl
and HTTP_RPC_PORT=50052
, an example output of a skaffold dev
execution on our getting-started example
curl localhost:50052/v1/events
{"result":{"timestamp":"2019-10-16T18:26:11.385251549Z","event":{"metaEvent":{"entry":"Starting Skaffold: {Version:v0.39.0-16-g5bb7c9e0 ConfigVersion:skaffold/v1 GitVersion: GitCommit:5bb7c9e078e4d522a5ffc42a2f1274fd17d75902 GitTreeState:dirty BuildDate:2019-10-03T15:01:29Z GoVersion:go1.13rc1 Compiler:gc Platform:linux/amd64}"}}}}
{"result":{"timestamp":"2019-10-16T18:26:11.436231589Z","event":{"buildEvent":{"artifact":"gcr.io/k8s-skaffold/skaffold-example","status":"In Progress"}},"entry":"Build started for artifact gcr.io/k8s-skaffold/skaffold-example"}}
{"result":{"timestamp":"2019-10-16T18:26:12.010124246Z","event":{"buildEvent":{"artifact":"gcr.io/k8s-skaffold/skaffold-example","status":"Complete"}},"entry":"Build completed for artifact gcr.io/k8s-skaffold/skaffold-example"}}
{"result":{"timestamp":"2019-10-16T18:26:12.391721823Z","event":{"deployEvent":{"status":"In Progress"}},"entry":"Deploy started"}}
{"result":{"timestamp":"2019-10-16T18:26:12.847239740Z","event":{"deployEvent":{"status":"Complete"}},"entry":"Deploy complete"}}
..
To get events from the API using gRPC
, first create a gRPC
client.
then, call the client.Events()
method:
func main() {
ctx, ctxCancel := context.WithCancel(context.Background())
defer ctxCancel()
// `client` is a gRPC client with connection to localhost:50051.
logStream, err := client.Events(ctx, &empty.Empty{})
if err != nil {
log.Fatalf("could not get events: %v", err)
}
for {
entry, err := logStream.Recv()
if err == io.EOF {
break
}
if err != nil {
log.Fatal(err)
}
log.Println(entry)
}
}
Each Entry in the log contains an Event in the LogEntry.Event
field and
a string description of the event in LogEntry.entry
field.
State API
The State API provides a snapshot of the current state of the following components:
- build state per artifacts
- deploy state
- file sync state
- status check state per resource
- port-forwarded resources
State API Contract
protocol | endpoint | encoding |
---|---|---|
HTTP | http://localhost:{HTTP_RPC_PORT}/v1/state |
newline separated JSON using chunk transfer encoding over HTTP |
gRPC | client.GetState(ctx) method on the SkaffoldService |
protobuf 3 over HTTP |
Examples
Using curl
and HTTP_RPC_PORT=50052
, an example output of a skaffold dev
execution on our microservices example
curl localhost:50052/v1/state | jq
{
"buildState": {
"artifacts": {
"gcr.io/k8s-skaffold/leeroy-app": "Complete",
"gcr.io/k8s-skaffold/leeroy-web": "Complete"
}
},
"deployState": {
"status": "Complete"
},
"forwardedPorts": {
"9000": {
"localPort": 9000,
"remotePort": 8080,
"namespace": "default",
"resourceType": "deployment",
"resourceName": "leeroy-web"
},
"50055": {
"localPort": 50055,
"remotePort": 50051,
"namespace": "default",
"resourceType": "service",
"resourceName": "leeroy-app"
}
},
"statusCheckState": {
"status": "Succeeded"
},
"fileSyncState": {
"status": "Not Started"
}
}
To retrieve the state from the server using gRPC
, first create gRPC
client.
Then, call the client.GetState()
method:
func main() {
// Create a gRPC client connection to localhost:50051.
// See code above
ctx, ctxCancel := context.WithCancel(context.Background())
defer ctxCancel()
grpcState, err = client.GetState(ctx, &empty.Empty{})
...
}
Control API
By default, skaffold dev
will automatically build artifacts, deploy manifests and sync files on every source code change.
However, this behavior can be paused and individual actions can be gated off by user input through the Control API.
With this API, users can tell Skaffold to wait for user input before performing any of these actions,
even if the requisite files were changed on the filesystem. By doing so, users can “queue up” changes while
they are iterating locally, and then have Skaffold rebuild and redeploy only when asked. This can be very
useful when builds are happening more frequently than desired, when builds or deploys take a long time or
are otherwise very costly, or when users want to integrate other tools with skaffold dev
.
The automation can be turned off or on using the Control API, or with auto-build
flag for building, auto-deploy
flag for deploys, and the auto-sync
flag for file sync.
If automation is turned off for a phase, Skaffold will wait for a request to the Control API before executing the associated action.
Each time a request is sent to the Control API by the user, the specified actions in the payload are executed immediately. This means that even if there are new file changes, Skaffold will wait for another user request before executing any of the given actions again.
Control API Contract
protocol | endpoint |
---|---|
HTTP, method: POST | http://localhost:{HTTP_RPC_PORT}/v1/execute , the Execution Service |
gRPC | client.Execute(ctx) method on the SkaffoldService |
HTTP, method: PUT | http://localhost:{HTTP_RPC_PORT}/v1/build/auto_execute , the Auto Build Service |
gRPC | client.AutoBuild(ctx) method on the SkaffoldService |
HTTP, method: PUT | http://localhost:{HTTP_RPC_PORT}/v1/sync/auto_execute , the Auto Sync Service |
gRPC | client.AutoSync(ctx) method on the SkaffoldService |
HTTP, method: PUT | http://localhost:{HTTP_RPC_PORT}/v1/deploy/auto_execute , the Auto Deploy Service |
gRPC | client.AutoDeploy(ctx) method on the SkaffoldService |
Examples
Using our Quickstart example, we can start skaffold with skaffold dev --auto-build=false
.
When we change main.go
, Skaffold will notice file changes but will not rebuild the image until it receives a request to the Control API with {"build": true}
:
curl -X POST http://localhost:50052/v1/execute -d '{"build": true}'
At this point, Skaffold will wait to deploy the newly built image until we invoke the Control API with {"deploy": true}
:
curl -X POST http://localhost:50052/v1/execute -d '{"deploy": true}'
These steps can also be combined into a single request:
curl -X POST http://localhost:50052/v1/execute -d '{"build": true, "deploy": true}'
We can make Skaffold start noticing file changes automatically again by issuing the requests:
curl -X PUT http://localhost:50052/v1/build/auto_execute -d '{"enabled": true}'
curl -X PUT http://localhost:50052/v1/deploy/auto_execute -d '{"enabled": true}'
To access the Control API via the gRPC
, create gRPC
client as before.
Then, use the client.Execute()
method with the desired payload to trigger it once:
func main() {
ctx, ctxCancel := context.WithCancel(context.Background())
defer ctxCancel()
// `client` is the gRPC client with connection to localhost:50051.
_, err = client.Execute(ctx, &pb.UserIntentRequest{
Intent: &pb.Intent{
Build: true,
Sync: true,
Deploy: true,
},
})
if err != nil {
log.Fatalf("error when trying to execute phases: %v", err)
}
}
Use the client.AutoBuild()
,client.AutoSync()
and client.AutoDeploy()
method to enable or disable auto build, auto sync and auto deploy:
func main() {
ctx, ctxCancel := context.WithCancel(context.Background())
defer ctxCancel()
// `client` is the gRPC client with connection to localhost:50051.
_, err = client.AutoBuild(ctx, &pb.TriggerRequest{
State: &pb.TriggerState{
Val: &pb.TriggerState_Enabled{
Enabled: true,
},
},
}) if err != nil {
log.Fatalf("error when trying to auto trigger phases: %v", err)
}
}