package main
import (
"bitbucket.org/afg/go-lambda-io/apigatewayproxyevent"
"github.com/eawsy/aws-lambda-go-core/service/lambda/runtime"
)
const (
ROUTE_BILLING_LIST = "list"
)
var HandleApiGateway = apigatewayproxyevent.Middleware(ApiGatewayHandle)
func ApiGatewayHandle(event *apigatewayproxyevent.Event, ctx *runtime.Context) (interface{}, error) {
switch event.Route {
case ROUTE_BILLING_LIST:
return BillingController(event)
default:
return event.Error(apigatewayproxyevent.NewErrRouteNotSupported(event.Route))
}
}
where
"bitbucket.org/afg/go-lambda-io/apigatewayproxyevent" is a private library of mine that I use to map and marshal events (like you do here https://github.com/eawsy/aws-lambda-go-event but I needed some custom manipulation so I created my own marshallers for different events), and the Middleware is just a passthrough function where I use defer in order to catch any un-handled situation and handle telemetry before exit, in this way the may function ApiGatewayHandle remains clean. Everything worked till today, when I added the middleware, could be the culprit?
// Middleware ...
func Middleware(handler func(event *Event, ctx *runtime.Context) (interface{}, error)) func(event *Event, ctx *runtime.Context) (interface{}, error) {
// deferred
return func(event *Event, ctx *runtime.Context) (interface{}, error) {
setTelemetryClient(event, ctx)
defer func() {
r := recover()
if r != nil { // in case of crash
switch t := r.(type) {
case string:
err := errors.New(t)
log.Println("is string")
log.Println(err)
case error:
// debug.PrintStack()
err := t
log.Println("is error")
log.Println(err)
log.Println(string(debug.Stack()))
default:
err := errors.New("unknown error")
log.Println(err)
}
} else { // run everytime before to terminate the process
logger.LogJSON(event, "deferred event")
event.Telemetry.TrackRequest(event.Env["AWS_LAMBDA_FUNCTION_NAME"], event.HTTPMethod, event.Path, event.RequestContext.RequestID, event.RequestStart, event.Duration(), event.Response.Code, event.IsSuccess())
event.Telemetry.Channel().Flush()
}
}()
return handler(event, ctx)
}
}
will try few other tests to figure out, thanks a lot in advance for any help you are able to give, thanks
did you saw this error comming up after a new deployement of your lambda ?
TelemetryClient is a 3rd party lib ?
if yes, please share the github repo
sure thing, this is the telemetry library I am trying to use
https://github.com/Microsoft/ApplicationInsights-Go
and this is the output given by the command:
2017-10-05T07:09:02Z
I can see they do use channels heavily
yes the error came after a new deployment
@riccardo-89 update your docker image so you have the latest version (some bugs of golang have been fixed in the last images). And for the go routine parts, if goroutines are not finished before the lambda returns then unexpected things can happen. So yes, see in your side and we are here if you need any further help. If your problem persists we may come with a private group to exchange more info and assist you more effectively.
@fsenart will do! yes you are right it should be something related to unfinished task at this point since I can see not all the telemetry are tracked correctly, thanks a lot for you disponibility and help, you have already done a lot building this lib I would feel guilty to ask for more of your time, really appreciated the thought, thanks :)
will have a look asap
I'm merging the other one ;)
@lsuss should be sorted :)
Instead of looking for
package.zip
ENOENT: no such file or directory, open 'package.zip'
my zip file is called
handler.zip
Is there something I need to do on the docker side?
I updated the top 2 lines to look the ones in the main go-aws-shim repo
HANDLER ?= handler
PACKAGE ?= $(HANDLER)
PACKAGE ?= $(HANDLER)
Although that seems to have broke the building process.
