Unable to run sync route execution longer than 60 seconds

[denismakogon]

Container unexpectedly dies during sync execution but runs fine in docker using command after 1 minute execution:

docker run -d -i -P denismakogon/aioservice-examples:latest

Images:

• denismakogon/aioservice-examples:latest
• denismakogon/sleep120:latest

Steps to reproduce:

curl -X POST -d '{"app":{"name":"testapp"}}' localhost:8080/v1/apps

curl -X POST -d '{"route":{"type": "sync", "path": "/sync", "image": "denismakogon/sleep120"}}' localhost:8080/v1/apps/testapp/routes

curl -X POST localhost:8080/r/testapp/sync

Function logs:
When POST to /r/{app}/route submitted

INFO[1442]                                               name=run.testapp.requests type=count value=1
INFO[1442]                                               name=run.testapp.waittime type=time value=0s

When container dies

INFO[1502] container status                              container_error= container_finished=2016-11-24 09:37:59.410257982 +0000 UTC container_running=false container_status=exited exit_code=137
INFO[1502]                                               name=run.testapp.error type=count value=1
INFO[1502]                                               name=run.testapp.time type=time value=1m0.06911744s
INFO[1502]                                               name="run.exec_time" type=time value=1m0.06911744s

But both containers are capable to run on Docker successfully.

It appears that sync func can run no longer than 60 seconds (Docker ReST client default HTTP timeout is 60 seconds).

Functions latest commit: af14904
Go version: 1.7
Docker version: 1.12.3, build 6b644ec

[ucirello]

This is a bug. The problem is that there is a hard-coded 30s timeout in the source code.

functions/api/server/runner.go

Line 172 in c14bc32

Timeout: 30 * time.Second,

@treeder @pedronasser @seiflotfy - what would the appropriate fix? Is the timeout an attribute of the route or the task call?

[denismakogon]

@ccirello would be the right fix to set execution timeout through API? There are three cases possible:

1. No timeout specified - use default 60 seconds (or whatever it be)
2. Timeout specified - run within timeout
3. Run until not finished

cc @treeder @pedronasser @seiflotfy

[ucirello]

would be the right fix to set execution timeout through API?

I don't see any other place to land the fix. :) It must be in the API - the only you way you can interface with IronFunctions daemon.

However, is this an attribute of the route (PUT /v1/apps/{app}/routes/{route}) or the task (POST /r/{app}/{route})?

[denismakogon]

@ccirello as i can see, it should be a part of routers API (POST /v1/apps/{app}/routes/{route}). Because in route payload we explicitly saying which functions we want to run along with route config instance, so it is a right place.

[pedronasser]

I think the Timeout should be configured in the route/apps configuration and even maybe via a Header. If no timeout is specified use the default. (like @denismakogon suggested)

[denismakogon]

@pedronasser to what kind of headers you're referring to?. We already have config object in payload, can't we just check on the API side if timeout was specified or not. In such case we have very small API impact (AFICS, only doc/swagger impact).