Mongrel users - Jun 2006 - Mongrel_rails restart and sudo

Sorry for the newbie question but I''m stumped:

I have an app that requires sudo to run mongrel_rails.  For my capistrano
recipe is do it like this:

sudo "mongrel_rails start -e production -p #{application_port} -d -c
#{current_path}"

Which works famously.

Now I have tried several things for my restart task.  Here they are:

sudo "cd #{current_path} && sudo mongrel_rails restart"

That doesn''t work.  For some reason the "cd..." command does
not work.  I''m
assuming that is some sort of problem with capistrano or my server.  So
moving on I tried:

sudo "mongrel_rails restart -c #{current_path}"

to avoid having to do the "cd...".  When I do that I get this in my
mongrel.log:

** USR2 signal received.
** Restarting with arguments:  ruby
/opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails start -e
production -p 81 -d -c /Library/WebServer/rails/kd/current
/opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails:158:in
`exec'': Operation not supported - ruby
/opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails start -e
production -p 81 -d -c /Library/WebServer/rails/kd/current
(Errno::EOPNOTSUPP)
        from 
/opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails:158:in
`run''
        from 
/opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/lib/mongrel/command.rb:188:
in `run''
        from 
/opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails:247
        from /opt/local/bin/mongrel_rails:18

Which I assume is due to the "sudo" not being added to the
"mongrel_rails
start..." command that is being called by "mongrel_rails
restart".  So it
seems that one cannot do "mongrel_rails restart" with
"sudo".  My ugly hack
is to to "mongrel_rails stop" and then "mongrel_rails
start".  Any ideas?

Now, I have control over this server and am fine with changing permissions
so that I don''t have to use "sudo".  This is probably
off-topic for this
list (better suited to "Server Admin for Dummies" perhaps) but how do
I do
that?  I assume I need sudo because mongrel needs to write stuff to a tmp
dir that my user does not have access to.  Any pointers?

Thanks,
Tom

"cd" does not respond to the sudo command. So separate the command  
into two lines, like so:

run "cd #{current_path}"
sudo "mongrel_rails restart"

--Alison
http://blog.alisonrowland.com

On Jun 16, 2006, at 11:25 AM, Tom Brice wrote:
> Sorry for the newbie question but I''m stumped:
>
> I have an app that requires sudo to run mongrel_rails.  For my  
> capistrano
> recipe is do it like this:
>
> sudo "mongrel_rails start -e production -p #{application_port} -d -c
> #{current_path}"
>
> Which works famously.
>
> Now I have tried several things for my restart task.  Here they are:
>
> sudo "cd #{current_path} && sudo mongrel_rails restart"
>
> That doesn''t work.  For some reason the "cd..." command
does not
> work.  I''m
> assuming that is some sort of problem with capistrano or my  
> server.  So
> moving on I tried:
>
> sudo "mongrel_rails restart -c #{current_path}"
>
> to avoid having to do the "cd...".  When I do that I get this in
my
> mongrel.log:
>
> ** USR2 signal received.
> ** Restarting with arguments:  ruby
> /opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails  
> start -e
> production -p 81 -d -c /Library/WebServer/rails/kd/current
> /opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails: 
> 158:in
> `exec'': Operation not supported - ruby
> /opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails  
> start -e
> production -p 81 -d -c /Library/WebServer/rails/kd/current
> (Errno::EOPNOTSUPP)
>         from
> /opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails: 
> 158:in
> `run''
>         from
> /opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/lib/mongrel/ 
> command.rb:188:
> in `run''
>         from
> /opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails:247
>         from /opt/local/bin/mongrel_rails:18
>
> Which I assume is due to the "sudo" not being added to the  
> "mongrel_rails
> start..." command that is being called by "mongrel_rails
restart".
> So it
> seems that one cannot do "mongrel_rails restart" with
"sudo".  My
> ugly hack
> is to to "mongrel_rails stop" and then "mongrel_rails
start".  Any
> ideas?
>
> Now, I have control over this server and am fine with changing  
> permissions
> so that I don''t have to use "sudo".  This is probably
off-topic for
> this
> list (better suited to "Server Admin for Dummies" perhaps) but
how
> do I do
> that?  I assume I need sudo because mongrel needs to write stuff to  
> a tmp
> dir that my user does not have access to.  Any pointers?
>
> Thanks,
> Tom
>
>
> _______________________________________________
> Mongrel-users mailing list
> Mongrel-users at rubyforge.org
> http://rubyforge.org/mailman/listinfo/mongrel-users

On 6/16/06 10:29 AM, Alison Rowland  wrote:
> two lines, like so:
> 
> run "cd #{current_path}"
> sudo "mongrel_rails restart"
Thanks Alison.

I though I tried this but I just rechecked.  Problem is capistrano.  When I
try this the cd executes but doesn''t "hold".  This means that
my
mongrel_rails restart command is executed in my home dir (not my
RAILS_ROOT).  It doesn''t find the PID file so it fails.

Any other ideas?

Tom

On 6/16/06 10:47 AM, Tom Brice  wrote:
> On 6/16/06 10:29 AM, Alison Rowland  wrote:
> 
>> two lines, like so:
>> 
>> run "cd #{current_path}"
>> sudo "mongrel_rails restart"
> 
> Thanks Alison.
> 
> I though I tried this but I just rechecked.  Problem is capistrano.  When I
> try this the cd executes but doesn''t "hold".  This means
that my mongrel_rails
> restart command is executed in my home dir (not my RAILS_ROOT).  It
doesn''t
> find the PID file so it fails.
> 
> Any other ideas?
> 
> Tom
Also just checked doing this at the server (to take capistrano out of the
equation):

sudo mongrel_rails restart

Still gives me the failure because the mongrel_rails start command executed
by mongrel_rails restart does not "inherit" sudo privileges.  So same
error
trace as my original message.

Tom

On 6/16/06, Tom Brice <tom at onidle.com> wrote:
>
> Sorry for the newbie question but I''m stumped:
>
> sudo "cd #{current_path} && sudo mongrel_rails restart"
>
> That doesn''t work.

Not having worked with capistrano (yet!  I will better myself one day!) I
really shouldn''t advice you, but here goes:

''cd'', as you note, doesn''t "hold" since it
affects the process it runs in,
and I assume that capistrano''s sudo either splits your command in two,
running them in two different, consecutive processes, or just sends it
along, in which case the system''s sudo doesn''t handle
&& very well.

But I assume that you have the full power of Ruby in capistrano, so you
could try

Dir.chdir(current_path)
sudo "mongrel_rails restart"

making sure to chdir back if you need to.  If that doesn''t work, how
about
putting your commands in a shell script and running that:

-----------script.sh--------
#!/bin/sh
cd whatever
mongrel_rails restart
-----------end script.sh----

with a

sudo "/path/to/script.sh"

I haven''t tried it, but it should be worth a shot.

Good luck,

/David
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/mongrel-users/attachments/20060616/054e124f/attachment.htm

I don''t know what''s going on with your restart task failing.
You
should probably fix that before worrying about Capistrano, but in the  
event that you get it working, here''s a bit more Cap info.

I just looked at my deploy.rb, and it has a little extra bit of  
structure which may help. Here''s my whole restart task, which I  
originally grabbed from somebody''s snippets:

desc "Restart the web server"
task :restart, :roles => :app do
   begin
     run "cd #{current_path}"
     sudo "mongrel_cluster_ctl restart"
   rescue RuntimeError => e
     puts e
     puts "Probably not a big deal, so I''ll just keep
trucking..."
   end
end

Perhaps wrapping it in begin...rescue...end lets Capistrano know to  
execute it all as one thing.

--Alison
http://blog.alisonrowland.com

On Jun 16, 2006, at 11:47 AM, Tom Brice wrote:
> On 6/16/06 10:29 AM, Alison Rowland  wrote:
>
>> two lines, like so:
>>
>> run "cd #{current_path}"
>> sudo "mongrel_rails restart"
>
> Thanks Alison.
>
> I though I tried this but I just rechecked.  Problem is  
> capistrano.  When I
> try this the cd executes but doesn''t "hold".  This means
that my
> mongrel_rails restart command is executed in my home dir (not my
> RAILS_ROOT).  It doesn''t find the PID file so it fails.
>
> Any other ideas?
>
> Tom
>
>
> _______________________________________________
> Mongrel-users mailing list
> Mongrel-users at rubyforge.org
> http://rubyforge.org/mailman/listinfo/mongrel-users

The implementation of ''sudo'' in capistrano is a little flawed.
It ought to
wrap the command in a ''sh -c'' wrapper to make sure that the
sudo privileges
apply to the entire command string and not just the first command. At the
moment you have to do that manually.

To do what you want do this

sudo <<-CMD
  sh -c "cd #{current_path} && mongrel_rails start"
CMD


The other problem is that capistrano is a little quick of the mark when
shutting down a connection - often before the last program has had a chance
to catch HUP signals. If this happens use.

sudo <<-CMD
  sh-c "cd #{current_path} && nohup mongrel_rails start
>/dev/null"
CMD

Rgs

NeilW
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/mongrel-users/attachments/20060616/fc8ffc24/attachment.htm

On 6/16/06 11:20 AM, Neil Wilson  wrote:
> To do what you want do this
> 
> sudo <<-CMD
>   sh -c "cd #{current_path} && mongrel_rails start"
> CMD
Thanks.  This still has the same problem with sudo not applying to the
mongrel_rails start command that mongrel_rails restart sends after sending
the USR2 signal.

I''m having other difficulties with Apache and file uploads. 
I''m gonna try
mongrel_cluster to see how that works.

Thanks to everyone for the suggestions.  At least I learned something!

Tom

On Fri, 2006-06-16 at 16:20 -0500, Tom Brice wrote:
> On 6/16/06 11:20 AM, Neil Wilson  wrote:
> 
> > To do what you want do this
> > 
> > sudo <<-CMD
> >   sh -c "cd #{current_path} && mongrel_rails start"
> > CMD
> 
> Thanks.  This still has the same problem with sudo not applying to the
> mongrel_rails start command that mongrel_rails restart sends after sending
> the USR2 signal.
> 
> I''m having other difficulties with Apache and file uploads. 
I''m gonna try
> mongrel_cluster to see how that works.
> 
> Thanks to everyone for the suggestions.  At least I learned something!
Tom, mongrel_cluster is the way to go for deployments with capistrano.
If you can post the stack dump you get in your mongrel.log after you do
this restart then I can probably see why it''s failing.


-- 
Zed A. Shaw
http://www.zedshaw.com/
http://mongrel.rubyforge.org/

On 6/16/06 4:40 PM, Zed Shaw  wrote:
> Tom, mongrel_cluster is the way to go for deployments with capistrano.
> If you can post the stack dump you get in your mongrel.log after you do
> this restart then I can probably see why it''s failing.
Here''s what I get:

** USR2 signal received.
** Restarting with arguments:  ruby
/opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails start -e
production -p 81 -d -c /Library/WebServer/rails/kd/current
/opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails:158:in
`exec'': Operation not supported - ruby
/opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails start -e
production -p 81 -d -c /Library/WebServer/rails/kd/current
(Errno::EOPNOTSUPP)
        from 
/opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails:158:in
`run''
        from 
/opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/lib/mongrel/command.rb:188:
in `run''
        from 
/opt/local/lib/ruby/gems/1.8/gems/mongrel-0.3.13/bin/mongrel_rails:247
        from /opt/local/bin/mongrel_rails:18

I''m gonna try to do mongrel_cluster right now.

Thanks,
Tom

I don''t get the same problem on Ubuntu linux using the same commands.
Are
you using Solaris?

Looks to me like exec is not holding root identity, which may mean that sudo
is only setting the Effective UID and not the real one.

Can you try this

sudo ruby -e ''puts Process.euid; puts Process.uid''

and see if you get two zeroes.

NeilW


On 16/06/06, Tom Brice <tom at onidle.com> wrote:
>
> On 6/16/06 4:40 PM, Zed Shaw  wrote:
>
> > Tom, mongrel_cluster is the way to go for deployments with capistrano.
> > If you can post the stack dump you get in your mongrel.log after you
do
> > this restart then I can probably see why it''s failing.
>
> Here''s what I get:
>

-- 
Neil Wilson (neil at aldur.co.uk)
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/mongrel-users/attachments/20060617/ebc3245b/attachment.htm

And while you''re at it try this as well.

sudo ruby -e ''exec "ruby -e \"puts Process.euid; puts
Process.uid\""''

NeilW
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://rubyforge.org/pipermail/mongrel-users/attachments/20060617/3b292928/attachment.htm

On 6/17/06 1:30 AM, Neil Wilson  wrote:
> Can you try this 
> 
> sudo ruby -e ''puts Process.euid; puts Process.uid''
> 
> and see if you get two zeroes.
SSH''d into my server I get this:
% sudo ruby -e ''puts Process.euid; puts Process.uid''
Password:
0
0


On 6/17/06 1:35 AM, Neil Wilson  wrote:
> And while you''re at it try this as well.
> 
> sudo ruby -e ''exec "ruby -e \"puts Process.euid; puts
Process.uid\""''
> 
> NeilW

I get:
% sudo ruby -e ''exec "ruby -e \"puts Process.euid; puts
Process.uid\""''
0
0

I''m now using mongrel_cluster and it''s cap tasks are starting
the server as
my user (thus not requiring sudo).  The mongrel_cluster restart seems to
work as well.

I''m still a bit baffled about the sudo on this machine.  I had to set
:use_sudo, FALSE for cap -a configure_mongrel_cluster.  The server is a
Xserve running OS X Server 10.4.5.

Tom

Looks like its not permissions then. Thinking about it I''d have
expected
EPERM rather than EOPNOTSUPP.

Are you still getting the mongrel cluster to use port 81, or are the ports
now above 1024?

NeilW

On 6/18/06 1:22 PM, Neil Wilson  wrote:
> Are you still getting the mongrel cluster to use port 81, or are the ports
> now above 1024?
I''m using 4 mongrels starting at 8000 for the cluster.  The errors I
had
previously were one mongrel (mongrel_rails start ... No mongrel_cluster)
which required sudo.  Now that I''m using mongrel_cluster the mongrels
start
under my user and do not require sudo.

Seems like :use_sudo defaulting to TRUE in Capistrano did not work for some
commands on my server.  Particularly anything that involved more than one
command.  One-liners execute sudo as expected, mulitple commands do not.

Let me know if you have any suggestions as I''m keen to understand the
system.  That said, for now I''m running.

Thanks for your help,
Tom

Does the one mongrel work if you use -p 8000 rather than -p 81 and
''run''
rather than sudo (or setting use_sudo to false which has the same effect)?

Port 81 is a privileged port, and port 8000 isn''t you see.

NeilW