When a function with executor type newdeploy got created, Newdeploy is expected to create deployment/HPA/service for it and insert serviceEntry to the cache for later use. Once clients call the function, newdeploy returns the serviceEntry to the router. However, the log shows that the newdeploy was unable to find the entry and prints "Resource not found - key 'xxx' not found". The root cause is that the informer controller instead of processing items in parallel, it dispatches XXFunc to process items one by one. So if there is any problem during the creation of the kubernetes resource, it takes a longer time to process the next item and hence the serviceEntry was not inserted before clients call the function. This PR lets the newdeploy to process items in extra goroutines instead of blocking the process loop. It's a workaround to solve the problem above, we should consider using workqueue to solve it in the future.
250 lines
7.5 KiB
Go
250 lines
7.5 KiB
Go
/*
|
|
Copyright 2016 The Fission Authors.
|
|
|
|
Licensed under the Apache License, Version 2.0 (the "License");
|
|
you may not use this file except in compliance with the License.
|
|
You may obtain a copy of the License at
|
|
|
|
http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
Unless required by applicable law or agreed to in writing, software
|
|
distributed under the License is distributed on an "AS IS" BASIS,
|
|
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
See the License for the specific language governing permissions and
|
|
limitations under the License.
|
|
*/
|
|
|
|
package executor
|
|
|
|
import (
|
|
"context"
|
|
"fmt"
|
|
"net/http"
|
|
"strings"
|
|
"sync"
|
|
"time"
|
|
|
|
"github.com/dchest/uniuri"
|
|
"github.com/pkg/errors"
|
|
"github.com/prometheus/client_golang/prometheus/promhttp"
|
|
"go.uber.org/zap"
|
|
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
|
|
|
|
fv1 "github.com/fission/fission/pkg/apis/fission.io/v1"
|
|
"github.com/fission/fission/pkg/crd"
|
|
"github.com/fission/fission/pkg/executor/cms"
|
|
"github.com/fission/fission/pkg/executor/fscache"
|
|
"github.com/fission/fission/pkg/executor/newdeploy"
|
|
"github.com/fission/fission/pkg/executor/poolmgr"
|
|
"github.com/fission/fission/pkg/executor/reaper"
|
|
fetcherConfig "github.com/fission/fission/pkg/fetcher/config"
|
|
)
|
|
|
|
type (
|
|
Executor struct {
|
|
logger *zap.Logger
|
|
|
|
gpm *poolmgr.GenericPoolManager
|
|
ndm *newdeploy.NewDeploy
|
|
cms *cms.ConfigSecretController
|
|
|
|
fissionClient *crd.FissionClient
|
|
fsCache *fscache.FunctionServiceCache
|
|
|
|
requestChan chan *createFuncServiceRequest
|
|
fsCreateWg map[string]*sync.WaitGroup
|
|
}
|
|
createFuncServiceRequest struct {
|
|
ctx context.Context
|
|
funcMeta *metav1.ObjectMeta
|
|
respChan chan *createFuncServiceResponse
|
|
}
|
|
|
|
createFuncServiceResponse struct {
|
|
funcSvc *fscache.FuncSvc
|
|
err error
|
|
}
|
|
)
|
|
|
|
func MakeExecutor(logger *zap.Logger, gpm *poolmgr.GenericPoolManager, ndm *newdeploy.NewDeploy, cms *cms.ConfigSecretController, fissionClient *crd.FissionClient, fsCache *fscache.FunctionServiceCache) *Executor {
|
|
executor := &Executor{
|
|
logger: logger.Named("executor"),
|
|
gpm: gpm,
|
|
ndm: ndm,
|
|
cms: cms,
|
|
fissionClient: fissionClient,
|
|
fsCache: fsCache,
|
|
|
|
requestChan: make(chan *createFuncServiceRequest),
|
|
fsCreateWg: make(map[string]*sync.WaitGroup),
|
|
}
|
|
go executor.serveCreateFuncServices()
|
|
|
|
return executor
|
|
}
|
|
|
|
// All non-cached function service requests go through this goroutine
|
|
// serially. It parallelizes requests for different functions, and
|
|
// ensures that for a given function, only one request causes a pod to
|
|
// get specialized. In other words, it ensures that when there's an
|
|
// ongoing request for a certain function, all other requests wait for
|
|
// that request to complete.
|
|
func (executor *Executor) serveCreateFuncServices() {
|
|
for {
|
|
req := <-executor.requestChan
|
|
m := req.funcMeta
|
|
|
|
// Cache miss -- is this first one to request the func?
|
|
wg, found := executor.fsCreateWg[crd.CacheKey(m)]
|
|
if !found {
|
|
// create a waitgroup for other requests for
|
|
// the same function to wait on
|
|
wg := &sync.WaitGroup{}
|
|
wg.Add(1)
|
|
executor.fsCreateWg[crd.CacheKey(m)] = wg
|
|
|
|
// launch a goroutine for each request, to parallelize
|
|
// the specialization of different functions
|
|
go func() {
|
|
fsvc, err := executor.createServiceForFunction(req.ctx, m)
|
|
req.respChan <- &createFuncServiceResponse{
|
|
funcSvc: fsvc,
|
|
err: err,
|
|
}
|
|
delete(executor.fsCreateWg, crd.CacheKey(m))
|
|
wg.Done()
|
|
}()
|
|
} else {
|
|
// There's an existing request for this function, wait for it to finish
|
|
go func() {
|
|
executor.logger.Debug("waiting for concurrent request for the same function",
|
|
zap.Any("function", m))
|
|
wg.Wait()
|
|
|
|
// get the function service from the cache
|
|
fsvc, err := executor.fsCache.GetByFunction(m)
|
|
|
|
// fsCache return error when the entry does not exist/expire.
|
|
// It normally happened if there are multiple requests are
|
|
// waiting for the same function and executor failed to cre-
|
|
// ate service for function.
|
|
err = errors.Wrapf(err, "error getting service for function %v in namespace %v", m.Name, m.Namespace)
|
|
req.respChan <- &createFuncServiceResponse{
|
|
funcSvc: fsvc,
|
|
err: err,
|
|
}
|
|
}()
|
|
}
|
|
}
|
|
}
|
|
|
|
func (executor *Executor) getFunctionExecutorType(meta *metav1.ObjectMeta) (fv1.ExecutorType, error) {
|
|
fn, err := executor.fissionClient.Functions(meta.Namespace).Get(meta.Name)
|
|
if err != nil {
|
|
return "", err
|
|
}
|
|
return fn.Spec.InvokeStrategy.ExecutionStrategy.ExecutorType, nil
|
|
}
|
|
|
|
func (executor *Executor) createServiceForFunction(ctx context.Context, meta *metav1.ObjectMeta) (*fscache.FuncSvc, error) {
|
|
executor.logger.Debug("no cached function service found, creating one",
|
|
zap.String("function_name", meta.Name),
|
|
zap.String("function_namespace", meta.Namespace))
|
|
|
|
executorType, err := executor.getFunctionExecutorType(meta)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
var fsvc *fscache.FuncSvc
|
|
var fsvcErr error
|
|
|
|
switch executorType {
|
|
case fv1.ExecutorTypeNewdeploy:
|
|
fsvc, fsvcErr = executor.ndm.GetFuncSvc(ctx, meta)
|
|
default:
|
|
fsvc, fsvcErr = executor.gpm.GetFuncSvc(ctx, meta)
|
|
}
|
|
|
|
if fsvcErr != nil {
|
|
e := "error creating service for function"
|
|
executor.logger.Error(e,
|
|
zap.Error(fsvcErr),
|
|
zap.String("function_name", meta.Name),
|
|
zap.String("function_namespace", meta.Namespace))
|
|
fsvcErr = errors.Wrap(fsvcErr, fmt.Sprintf("[%s] %s", meta.Name, e))
|
|
} else if fsvc != nil {
|
|
_, err = executor.fsCache.Add(*fsvc)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
}
|
|
|
|
executor.fsCache.IncreaseColdStarts(meta.Name, string(meta.UID))
|
|
|
|
return fsvc, fsvcErr
|
|
}
|
|
|
|
// isValidAddress invokes isValidService or isValidPod depending on the type of executor
|
|
func (executor *Executor) isValidAddress(fsvc *fscache.FuncSvc) bool {
|
|
if fsvc.Executor == fscache.NEWDEPLOY {
|
|
return executor.ndm.IsValid(fsvc)
|
|
} else {
|
|
return executor.gpm.IsValid(fsvc)
|
|
}
|
|
}
|
|
|
|
func serveMetric(logger *zap.Logger) {
|
|
// Expose the registered metrics via HTTP.
|
|
metricAddr := ":8080"
|
|
http.Handle("/metrics", promhttp.Handler())
|
|
err := http.ListenAndServe(metricAddr, nil)
|
|
|
|
logger.Fatal("done listening on metrics endpoint", zap.Error(err))
|
|
}
|
|
|
|
// StartExecutor Starts executor and the executor components such as Poolmgr,
|
|
// deploymgr and potential future executor types
|
|
func StartExecutor(logger *zap.Logger, functionNamespace string, envBuilderNamespace string, port int) error {
|
|
fissionClient, kubernetesClient, _, err := crd.MakeFissionClient()
|
|
if err != nil {
|
|
return errors.Wrap(err, "failed to get kubernetes client")
|
|
}
|
|
|
|
err = fissionClient.WaitForCRDs()
|
|
if err != nil {
|
|
return errors.Wrap(err, "error waiting for CRDs")
|
|
}
|
|
|
|
fetcherConfig, err := fetcherConfig.MakeFetcherConfig("/userfunc")
|
|
if err != nil {
|
|
return errors.Wrap(err, "Error making fetcher config")
|
|
}
|
|
|
|
restClient := fissionClient.GetCrdClient()
|
|
fsCache := fscache.MakeFunctionServiceCache(logger)
|
|
|
|
poolID := strings.ToLower(uniuri.NewLen(8))
|
|
reaper.CleanupOldExecutorObjects(logger, kubernetesClient, poolID)
|
|
go reaper.CleanupRoleBindings(logger, kubernetesClient, fissionClient, functionNamespace, envBuilderNamespace, time.Minute*30)
|
|
|
|
gpm := poolmgr.MakeGenericPoolManager(
|
|
logger,
|
|
fissionClient, kubernetesClient,
|
|
functionNamespace, fetcherConfig, poolID)
|
|
|
|
ndm := newdeploy.MakeNewDeploy(
|
|
logger,
|
|
fissionClient, kubernetesClient, restClient,
|
|
functionNamespace, fetcherConfig, poolID)
|
|
|
|
cms := cms.MakeConfigSecretController(logger, fissionClient, kubernetesClient, ndm, gpm)
|
|
|
|
api := MakeExecutor(logger, gpm, ndm, cms, fissionClient, fsCache)
|
|
|
|
go api.Serve(port)
|
|
go serveMetric(logger)
|
|
|
|
return nil
|
|
}
|