Provide steps to re-deploy after restarting Vagrant VM

[lfzCarlosC]

problem description

For Hello case
./wsk action invoke /whisk.system/samples/echo -p message hello --blocking --result

returns "error: [Errno 111] Connection refused"

steps to reproduce

1. Clone openwhisk
   git clone --depth=1 https://github.com/openwhisk/openwhisk.git
2. Change directory to tools/vagrant/simple
   cd openwhisk/tools/vagrant/simple
3. Run script to create vm and run hello action
   vagrant up
   ./wsk action invoke /whisk.system/samples/echo -p message hello --blocking --result

[lfzCarlosC]

I try the quickcase following the guide https://github.com/openwhisk/openwhisk

[lfzCarlosC]

I run it locally on my laptop

[csantanapr]

This reminds me to ask, @rabbah or @domdom82 do you know if Travis supports vagrant?

It will be awesome to have automate tests checking the vagrant use cases.

[csantanapr]

Hi @lfzCarlosC could you provide more info on your environment to be able to reproduce?

I tried today with latest hash bf3fe240ff56f772837f7c9ccf909483b1893a8c

I'm using Mac OSX El Capitan

$ uname -a
Darwin carlosmac.ibm.com 15.5.0 Darwin Kernel Version 15.5.0: Tue Apr 19 18:36:36 PDT 2016; root:xnu-3248.50.21~8/RELEASE_X86_64 x86_64

$ vagrant -v
Vagrant 1.8.1

Virtual Box Version 5.0.20 r106931

I clean ~/.vagrant.d/ just in case
I made sure no whisk VMs are listed or running on Virtual Box
No other VMs are running
Steps:

$ rm -rf ~/.vagrant.d/
$ git clone --depth=1 https://github.com/openwhisk/openwhisk.git
$ cd openwhisk/tools/vagrant/simple/
$ ./hello

wait for a while, downloading new vagrant ubuntu to cache (10 minutes) then whisk setup (20 minutes)

==> default: ++ bin/wsk action invoke /whisk.system/samples/echo -p message hello --blocking --result
==> default: {
==> default:     "message": "hello"
==> default: }
{
    "message": "hello"
}

$ git log
commit bf3fe240ff56f772837f7c9ccf909483b1893a8c
Author: Perry Cheng <perry@us.ibm.com>
Date:   Wed Jun 8 21:34:48 2016 -0400

[lfzCarlosC]

I clean all the things and re-ran the whole steps,This time the behaviour is a little bit different. Hello action work fine the first time when vagrant vm is created , but then if I try to restart vm(via GUI from virtual box) and run hello action , the error happens.I will attach the log info then

[domdom82]

@csantanapr I have run a travis build that does install a vagrant box. It would work in theory, however you need to install virtualbox first which in conjunction with our own build + test flow takes longer than travis allows on the free tier and the build gets killed. I suggest making a build on our Jenkins that does a vagrant up and checks the output.

[csantanapr]

@lfzCarlosC Ah I think you are hitting some of the UX, we are seeing and working on fixing.
Restarting I think will cause problems, since you have to deploy ant deploy but the inmortal DB is not up. @houshengbo is working on fixing this workflow.

For now don't restart the VM, just suspend it and then up (not resume). you can use vagrant cli, we have that in the [simple/README.md(https://github.com/openwhisk/openwhisk/blob/master/tools/vagrant/simple/README.md)

# Suspend vagrant vm when done having fun
  vagrant suspend

# Resume vagrant vm to have fun again
  vagrant up

[csantanapr]

@domdom82 thanks for volunteering 😆

I suggest making a build on our Jenkins that does a vagrant up and checks the output.

I would prefer to see something externally test this, but I guess for now we will use an internal server, but even with that I will suggest to have all the scripts in open (i.e. tools/travis/test-vagrant.sh)

[sjfink]

I don't think you can run vagrant inside an SL VM. Virtualization inside virtualization usually doesn't work.

[domdom82]

relate issue #311

[lfzCarlosC]

But I was considering a real development case.If I developed new updates for action/trigger/rule and test it on local vagrant env, next time these updates are not useful because I am not allowed to restart the vm? Or what is the process of development based on local vagrant env?

[rabbah]

I don't understand the last comment. If you restart the vm you expect the whisk deployment to still be up?

[lfzCarlosC]

let us consider a case. If I develop an action A on vagrant vm from my host machine, and then the vm for some reasons dump ,so how should I do to make action A rerun?

[csantanapr]

Suspending the machine is like shutting it down. It doesn't take resources. Just suspend , when you start the VM your actions, rules, etc are still available.

We also have a custom setup where you can use a persistent remote DB like cloudant or a remote CouchDB

Regardless if you restart the VM the whisk system doesn't start automatically, you need to deploy/start the whisk system.

[lfzCarlosC]

So, can I use this guide https://developer.ibm.com/open/2016/02/22/openwhisk-a-quick-tech-preview/ if I want to deploy whisk system once I need?

[domdom82]

Regardless if you restart the VM the whisk system doesn't start automatically, you need to deploy/start the whisk system.

@csantanapr not entirely true. The system should come back up in general. Ephemeral couch is a different story, however. If you use cloudant or other peristent db it should recover from a reboot. Part of our RAS strategy.

[lfzCarlosC]

@csantanapr I clean up my history cache file and re run today. But a different error happens.This is the link

#624

[csantanapr]

I'm working on providing instructions on how to re-deploy openwhisk after restarting the Vagrant VM.

[csantanapr]

This is fix now instructions in vagrant/README.md