gmane.comp.lang.ruby.capistrano.general

Re: Deploying to servers behind a gateway

Irish — 2012-05-24T18:34:17

Hey Cory,
    I have to implement a similiar setup to yours, where theres a gateway 
server between me and the deployment server.  Did you ever get this working?



On Monday, April 16, 2012 6:34:29 PM UTC-7, Cory Foy wrote:

Re: Err or during cap deploy:update

Chris Roberts — 2012-05-23T12:50:02

This worked thank you.

On Wed, May 23, 2012 at 1:04 AM, Donovan Bray <donnoman-Re5JQEeQqe8AvxtiuMwx3w< at >public.gmane.org> wrote:

Re: Rails App doesn't find gems after deploy

Lee Hambley — 2012-05-23T06:43:51

For what it's worth, I absolutely agree with Donovan in this respect, that said I do use Chef provisioned environments with rbenv, chiefly because rbenv does a better job of installing Ruby and it's obscure dependencies (should we rather say, unintelligent searching of dependencies) and in order to create a layer of separation between my provisioning environment, and my application environment.  

All of that being said, I know the risks, I know where to look when things go wrong, and I know that even then, it's a bad idea. (But Chef doesn't make it that easy)

As is almost always the case, people are misunderstanding the way Capistrano works, and relying on dotfile hacks, and shell environment modification scripts (rvm) and wondering why it doesn't work in an environment with no shell and which doesn't source dotfiles.

Rbenv at least provides `local-ruby-exec` which is a replacement for the `#!/bin/env ruby` line, which takes care of the modifications to the environment, this at least makes it tolerable in production, but requires modifications to the PATH, to Capistrano, to the bundle configuration and more, for me this is a worthy trade-off.

I would advise against doing anything with RVM in production, period. If you must (you mustn't) use something like this in production look at rbenv; it's from a 37signals employee, and it's particularly well written. (Don't do it.)

If you are using Chef ( as I am ) and you want to separate Chef from your application, don't - modify your bootstrap to install the system version of Ruby that you desire (I wish I would have done this, but in my position as Cap maintainer I owe it to the community to experiment with various configurations)

- Lee  


On Tuesday, May 22, 2012 at 10:28 PM, Ramsés López wrote:

Re: Rails App doesn't find gems after deploy

Donovan Bray — 2012-05-23T05:22:42

I'm going to get up on my soapbox a bit. I believe requiring more than the system version of ruby is a systems operations smell. Your violating some very basic separation of concerns by running dissimilar applications on the same box, and that way lies pain. You are likely creating an all eggs in one basket situation. 

Rvm will only make your life harder and troubleshooting production much more difficult and fragile. Not dissing rvm here; any solution that purports to solve this problem muddies the waters and I don't believe belongs on production. If you want ruby isolation; then do real isolation use more boxes. 

With virtualization you can insulate systems on physical hardware; if you are using a cloud you can use more but smaller instances. If your already on the smallest instance your cloud provider provides and your putting multiple apps on it; neither one is going to perform well and your cutting your nose off to spite your face. Plan for success and give each app a chance to breathe and succeed on their own. 

There's rarely a good argument for trying to run multiple versions of ruby on the same production system.  

On your Dev env knock yourself out; I use rvm with vagrant and rvm inside the vagrant.  But it doesn't belong on production systems. 


On May 22, 2012, at 1:28 PM, Ramsés López <ramses.lopezs-Re5JQEeQqe8AvxtiuMwx3w< at >public.gmane.org> wrote:

Re: Err or during cap deploy:update

Donovan Bray — 2012-05-23T05:04:08

Use the git form of specifying the repo instead of http

See http://help.github.com/deploy-with-capistrano/

On May 22, 2012, at 12:03 PM, opensourcechris <chris.gogreen-Re5JQEeQqe8AvxtiuMwx3w< at >public.gmane.org> wrote:

Re: Rails App doesn't find gems after deploy

Hassan Schroeder — 2012-05-22T22:52:06



Does `gem list --local` show you the correct gems? Can you start
up a console successfully while logged in?

If so, then your Rails app is probably being run as a different user,
without access to your RVM environment.

If not, run `bundle install` and try restarting your app.

HTH,

Err or during cap deploy:update

opensourcechris — 2012-05-22T19:03:25

I can not find what is wrong. I have tried to change the passwords
then try again with no luck. This is the error message I am recieving.

  * executing `deploy:update_code'
    executing locally: "git ls-remote https://chrisgogreen-9UaJU3cA/F/QT0dZR+AlfA< at >public.gmane.org/chrisgogreen/deckchainrails.git
HEAD"
Password for 'https://chrisgogreen-9UaJU3cA/F/QT0dZR+AlfA< at >public.gmane.org':
    command finished in 10205ms
  * executing "git clone -q https://chrisgogreen-9UaJU3cA/F/QT0dZR+AlfA< at >public.gmane.org/chrisgogreen/deckchainrails.git
/home/deployer/public/deckchain.com/public/deckchainrails/releases/
20120522190033 && cd /home/deployer/public/deckchain.
com/public/deckchainrails/releases/20120522190033 && git checkout -q -
b deploy 97cb1b474fa60f164155242a4beb96fb906f8906 && (echo
97cb1b474fa60f164155242a4beb96fb906f8906 > /home/deployer/public/
deckchain.com/publ
ic/deckchainrails/releases/20120522190033/REVISION)"
    servers: ["66.228.43.38"]
Password:
    [66.228.43.38] executing command
 ** [66.228.43.38 :: err] Password:
Password:


Please any help will be appreciated.

Rails App doesn't find gems after deploy

Ramsés López — 2012-05-22T20:28:26

Hello all, 

I'm trying to implement capistrano to deploy from my workstation to another 
server. I followed the instructions from 2.x-From-The-Beginning<https://github.com/capistrano/capistrano/wiki/2.x-From-The-Beginning>, 
and I configured my apache server according to this serverfault answer<http://serverfault.com/a/65416/116448>
.

However, when I use capistrano to transport it to the other server, the 
rails app doesn't seems to find methods containen in gems, like paginate 
from will_paginate. If I remove the references to that gem, then rails will 
complain about not finding some other method of another gem, like the ones 
from blueprint-rails.

I'm guessing that it could be some problem with the config of my 
environment, or maybe in my capistrano config, but right now I'm out of 
ideas, and don't know what else to check. Any ideas?

Some info that might help:

Ruby:

$ rvm list

rvm rubies

   ruby-1.9.2-p290 [ i686 ]
=* ruby-1.9.3-p194 [ i686 ]

Capistrano:
$ cap --version
Capistrano v2.12.0

My deploy.rb
http://pastie.org/3951922

My Capfile
http://pastie.org/3951928

Thanks! any suggestions will be appreciated

Re: deploying a branch

Donovan Bray — 2012-05-22T15:48:48

BRANCH=testing cap deploy

Then in your deploy.rb somewhere 

set :branch, ENV['BRANCH'] || 'master'

Change 'master' to be whatever branch you want to use when you don't specify anything. 



On May 21, 2012, at 5:57 PM, Benjamin Moses Lieb <red-2zkBS/BNfO7XOH51mAaJLw< at >public.gmane.org> wrote:

Re: How can I make `run` or `download` methods to always use the :roles I declared in the code?

Lee Hambley — 2012-05-22T07:58:28

Hi,

Unfortunately this is expected to behave differently for everyone, host or role filters should be additive, or subtractive; there is no "correct behaviour"

I encourage you to engineer your recipes to use servers_for_role() when you need to bypass filters that filter out too much, and to avoid engineering your recipes to rely on host/role specific configuration.


Configuration file management is better served with Chef, Puppet, BOSH or similar [1].  

If steps 1&3 were about making a "Deployment in progress" page, You can do that with try_files (NGinx directive), and upload a maintenance message, removing it when you are done, `cap web:disable` does this (and might be removed, depending which version you are using)

--
[1] http://en.wikipedia.org/wiki/Comparison_of_open_source_configuration_management_software



On Tuesday, May 22, 2012 at 9:52 AM, 赵 耀 wrote:

How can I make `run` or `download` methods to always use the :roles I declared in the code?

赵耀 — 2012-05-22T07:52:15

Hi everyone,

When I tried to do this:
1. remove the target server from my nginx upstream 
2. reload nginx
3. restart my app server #
4. add back to nginx upstream
5. reload nginx.

 Because I declared a ROLES or HOSTS env, all commands are running at the app server. The env hide the :roles => :nginx I passed into the `run` `download` methods.
For some reasons I can not declare :roles => on the task, which means I must use the env vars to declare which host I want to deploy.
How can do this?



Best regards.

Re: I'm getting this error when running this capistrano task (my own - cap deploy:show_pointers):

Lee Hambley — 2012-05-22T07:50:25

Benjamin,

It's just that if I know the question is at SO, I'll click the link, see if you accepted an answer, and I'll decide to ignore the mailing list post, or confirm that the SO answer was the sane one, when you don't disclose that it's cross-posted, potentially you're wasting the SO member community's time (you get the right answer here), or our time (you already got the right answer there)

I absolutely believe in using the power of SO to write a good question with code highlighting, explaining things in detail, and having a point of reference for your query, then nipping to a mailing list or IRC, or whatever - tell them the question comes with code samples, and logs/etc - and it's at SO…. I quite like that model, it means i can earn some SO rep and have something in return for all the time invested answering support queries :)

I don't want to chastise you too heavily, it's a slightly obscure point of view that I hold, but it encourages me to discontinue the Capistrano group, and close the GH issues tracker (not your fault, but 50% of all queries are "how does ssh work" / "my deployment doesn't work") -- community maintainership and support take up 80% of the time I can spend working on Capistrano, and if another community already had the chance to help people, it would free me to work on improvements.

- Lee  


On Tuesday, May 22, 2012 at 2:55 AM, Benjamin Moses Lieb wrote:

Re: how deploy a branch

Lee Hambley — 2012-05-22T07:45:15

Yes

On Tuesday, May 22, 2012 at 2:40 AM, Benjamin Moses Lieb wrote:

Re: I'm getting this error when running this capistrano task (my own - cap deploy:show_pointers):

Benjamin Moses Lieb — 2012-05-22T00:55:41

< at >lee 

I'm sorry to have offended you. I simply come here for a collaborative 
environment to problem solving. I really don't understand what you mean by 
being rude and sneaky. That was obviously not my intention. Next time I 
cross post here I'll try to remember to mention that I have done so. 

< at >dbray,

Thanks for the advice, I'm sure that's the issue. I've referenced you as 
having solved the enigma on SO.

deploying a branch

Benjamin Moses Lieb — 2012-05-22T00:57:49

 

I've read that with -s I can deploy a specific branch. I tried this and it 
failed. The branch is only on my local dev machine. Does it also have to be 
on the remote server?

(This post is not cross-posted anywhere)

how deploy a branch

Benjamin Moses Lieb — 2012-05-22T00:40:43

 

I've read that with -s I can deploy a specific branch. I tried this and it 
failed. The branch is only on my local dev machine. Does it also have to be 
on the remote server?

Re: error: no-auth-cache -rHEAD returned status code pid 65009 exit 1

Rob Hunter — 2012-05-21T15:48:00

What happens when you run that command yourself?

Any errors or interactive prompts?

Re: What's the difference between ssh remote command and capistrano run?

Rob Hunter — 2012-05-21T15:41:23

There are a few differences.

`cap run httpd` works very much like `ssh host httpd` but it differs in a
few ways:

It doesn't claim to be an interactive terminal. Anything that behaves
differently for interactive use and non-interactive use will use the
non-interactive behavior for Capistrano.

By default, `sudo` asks for passwords on interactive terminals but never
tries asking a non-terminal. Check your security logs to see if this is
happening to you.

Capistrano doesn't use the `ssh` command under the covers -- Capistrano
uses the Ruby library `Net::SSH` instead of the C OpenSSH code.

If your SSH configuration uses some of OpenSSH more advanced features (like
proxies or X11 forwarding) they may not work the same under Ruby's Net::SSH.

A useful debugging trick: compare the results of `ssh myhost env` against
the results of `cap run env`. That will show any differences in environment
variables.

Likewise, you can compare the results of `whoami` or `pwd`. If everything
is still the same, running something with `sudo`.

Good luck finding the difference!

Re: I'm getting this error when running this capistrano task (my own - cap deploy:show_pointers):

Lee Hambley — 2012-05-21T15:02:57

Donovan, no problem, I don't want to act too dogmatic (plus I'm having a serious case of the Mondays) but it really does get my goat that people don't think about this kind of stuff when x-posting.

30 seconds to write an email that what, 800 people on the list (mostly all silent) have to deal with then. 


On Monday, May 21, 2012 at 5:01 PM, Donovan Bray wrote:

Re: I'm getting this error when running this capistrano task (my own - cap deploy:show_pointers):

Donovan Bray — 2012-05-21T15:01:32

Lee np, I answered before I got to your post. I don't use so and had no idea. I lurk here and that's about it. 

On May 21, 2012, at 7:53 AM, Lee Hambley <lee.hambley-Re5JQEeQqe8AvxtiuMwx3w< at >public.gmane.org> wrote:

Re: I'm getting this error when running this capistrano task (my own - cap deploy:show_pointers):

Lee Hambley — 2012-05-21T14:53:39

Donovan,

I appreciate that you took the time to answer him, but I'd thank you not to reward cross-posters. 

I follow the capistrano tag on SO, and an RSS feed of the google search results, and Google notifications for the common search terms, and so much effort is wasted on duplicate effort from people cross-posting, it's quite frustrating.

Benjamin appears to be a freelancer, and especially when time literally is money, I can appreciate the desire for a speedy answer, but that is still no justification for cross-posting.

To be clear, finally, cross-posting with disclosure is fine, in my book; such as "the question worked better with decent code samples, so I posted it to SO, it's a little quiet over there, so I'm sharing a link here" would be absolutely fine.

- Lee 


On Monday, May 21, 2012 at 4:50 PM, Donovan Bray wrote: