application_spawner SEG FAULT about passenger HOT 19 CLOSED

From honglilai on May 01, 2008 09:50:44

Are you sure you don't have bad RAM or other bad hardware?

from passenger.

From drew.bomhof on May 01, 2008 10:51:23

I would be about 99.9% sure I don't have bad hardware. This box is slightly over a
year old and all other VM's including DOM0 are running without incident. I create my
Virtual Machines with minimal OS installs, so many times I find that obscure errors
have something to do with a missing dependency.

Thank you for your help in this matter.

from passenger.

From honglilai on May 01, 2008 11:09:40

What is the owner of the file /usr/local/rails_apps/analysis/config/environment.rb?

And could you comment out
/var/lib/gems/1.8/gems/passenger-1.0.3/lib/passenger/application_spawner.rb line 229,
restart Apache, and check whether that still crashes?

from passenger.

From drew.bomhof on May 01, 2008 11:26:56

environment.rb owner: root:root owner writable world readable

Ok cool! I commented out that line and now I can see my site.

Thanks a lot for all your help. I really appreciate it!

Interested to find out why it happened.

from passenger.

From honglilai on May 01, 2008 11:57:38

Yeah. It seems that changing the supplementary groups of a process will crash Ruby.

Could you run the following code, in irb, as root, and check whether it crashes?

require 'etc'
Process.initgroups("nobody", Etc.getpwnam("nobody").gid)

from passenger.

From drew.bomhof on May 01, 2008 12:03:39

Here are the results. It didn't crash.

irb(main):001:0> require 'etc'
=> true
irb(main):002:0> Process.initgroups("nobody",Etc.getpwnam("nobody").gid)
=> [546, 65534]
irb(main):003:0>

from passenger.

From honglilai on May 01, 2008 12:14:22

What happens if you do this?

• Uncomment the line you commented out earlier.
• Chown environment.rb to a normal user.

from passenger.

From drew.bomhof on May 01, 2008 12:35:41

No go.

I get the original SEG FAULT error again.

I tried with a couple different accounts: , www-data ...

from passenger.

From honglilai on May 01, 2008 12:42:14

Could you run the following code, in irb, as root, and check whether it crashes?

require 'etc'
g = Process.initgroups("nobody", Etc.getpwnam("nobody").gid)
Process.groups = g

from passenger.

From drew.bomhof on May 01, 2008 12:46:41

Still no crash.

root@develop:/usr/local/rails_apps/analysis# irb
irb(main):001:0> require 'etc'
=> true
irb(main):002:0> g = Process.initgroups("nobody", Etc.getpwnam("nobody").gid)
=> [546, 65534]
irb(main):003:0> Process.groups = g
=> [546, 65534]

from passenger.

From honglilai on May 01, 2008 13:12:16

Try this:

1. Change ownership of environment.rb back to root.

2. Edit application_spawner.rb line 229. Change:

   Process.groups = Process.initgroups(username, gid)

to

STDERR.puts "username = #{username}, gid = #{gid}"
STDERR.flush
Process.groups = Process.initgroups(username, gid)

What do you see in your Apache error log now when the spawn manager crashes?

from passenger.

From drew.bomhof on May 01, 2008 13:32:47

I think this result will help you out.

username = nobody,gid=65534

According to getent the gid should be 546

root@develop:/usr/local/rails_apps/analysis# getent group | grep nobody
Guests:*:546:nobody

from passenger.

From drew.bomhof on May 01, 2008 13:35:35

Ok, I think I have one of the issues. If I run getent passwd | grep nobody

nobody:x:65534:65534:nobody:/nonexistent:/bin/sh
nobody:x:999:514:nobody:/dev/null:/bin/false

I'm showing two nobody accounts. Hence the array thats showing up.

So I'm assuming if I fix the dual nobody's it might fix the problem.

from passenger.

From drew.bomhof on May 02, 2008 06:22:16

Removing the extra nobody user didn't fix the problem. But after further
investigation I'm wondering if it might have something to do with my LDAP
configuration. Either way commenting out line 229 in application_spawner.rb is
resolving the problem for now. Thanks for you all your help on this great product.

from passenger.

From honglilai on May 09, 2008 07:36:21

Hi Drew.

I'm wondering whether this problem would be solved if you use the development version
(from the git repository) in combination with conservative spawning. That is, by
setting "RailsSpawnMethod" to "conservative": http://www.modrails.com/documentation/Users%20guide%20latest.html#_railsspawnmethod_lt_string_gt Could you give it a try? Thanks.

from passenger.

From honglilai on May 19, 2008 05:05:13

Labels: NeedsConfirmation

from passenger.

From stanislav.bozhkov on June 10, 2008 15:40:16

I'm using Fedora 9 x86_64. Today I installed passenger-1.9.0 and also have SEG FAULT
from-time-to-time. However, I can't reproduce it.

from passenger.

From honglilai on June 11, 2008 01:44:51

stanislav.bozhkov: when the segfault happens, do you see any backtraces in the Apache
error log? What exactly segfaults?

from passenger.

From honglilai on July 16, 2008 08:37:31

I'm closing this bug because of lack of feedback.

Status: WontFix

from passenger.