/etc/hosts can prevent puma from starting

[AKovtunov]

After some coding got this error on running rails s:

puma-3.5.0/lib/puma/binder.rb:268:in `initialize': Address already in use - bind(2) for "127.0.0.1" port 3000 (Errno::EADDRINUSE)

My environment is:

$ rails -v       
Rails 5.0.0
$ ruby -v
ruby 2.3.1p112 (2016-04-26 revision 54768) [x86_64-linux]

I've tried:

1. Create a new project - the same
2. Checked Rails 4.2 - problem solved
3. Reinstall rails 5 and ruby - the same problem
4. lsof -wni tcp:3000 returns me nothing
5. ps aux | grep "rails" - nothing
6. ps aux | grep "puma" - nothing
7. ps aux | grep "ruby" -nothing
8. Use puma instead of rails s - problem solved
9. Use RAILS_ENV=production bundle exec rails s - problem solved
10. Use rails s -p 3001 - same problem, for other ports too

Monkey-patched https://github.com/puma/puma/blob/master/lib/puma/binder.rb and added pry.
Found that TCPServer.new is called few times with 127.0.0.1:3000

[3] pry(#<Puma::Binder>)> TCPSocket.gethostbyname "localhost"
=> ["localhost",
 [my hosts?],
 2,
 "127.0.0.1",
 "127.0.0.1",
 "127.0.0.1",
 "127.0.0.1",
 "127.0.0.1"]

I think this commit broke something for my machine 9b1de10#diff-4fd9276d9f86d7f4827e6ca4ab6b82b8
And I think this is related to the #782

[AKovtunov]

Downgrading to 3.4.0 works as charm

[evanphx]

Oh geez. Ok, I’ll fix this asap and have 3.5.1 out.

On Wed, Jul 20, 2016 at 2:46 AM Alexandr Kovtunov notifications@github.com
wrote:

Downgrading to 3.4.0 works as charm

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#1022 (comment), or mute
the thread
https://github.com/notifications/unsubscribe-auth/AAAAByheOt_CQU9hu8-pmogNxalForXIks5qXe6AgaJpZM4JQmLn
.

[evanphx]

@AKovtunov i'm curious, what does your /etc/hosts file look like? I'm wondering why localhost returned 127.0.0.1 multiple times..

[AKovtunov]

After that I was wondering the same too and cleared a bit my hosts file.
I was working on a project with multi-service architecture and needed to bind some routes to 127.0.0.1 :)

But the error is still here, here is my basic hosts that were installed by ubuntu

[1] pry(#<Puma::Binder>)> localhost_addresses
=> ["127.0.0.1", "127.0.0.1"]
[2] pry(#<Puma::Binder>)> TCPSocket.gethostbyname "localhost"
=> ["localhost", ["alexandr-Lenovo-IdeaPad-Y500", "localhost", "alexandr-Lenovo-IdeaPad-Y500.localhost"], 2, "127.0.0.1", "127.0.0.1"]

‹ruby-2.3.1›  ‹master*› $ cat /etc/hosts                                                                                                              1 ↵
127.0.0.1   localhost
127.0.1.1   alexandr-Lenovo-IdeaPad-Y500
127.0.0.1 alexandr-Lenovo-IdeaPad-Y500.localhost alexandr-Lenovo-IdeaPad-Y500 localhost
# The following lines are desirable for IPv6 capable hosts
::1     ip6-localhost ip6-loopback
fe00::0 ip6-localnet
ff00::0 ip6-mcastprefix
ff02::1 ip6-allnodes
ff02::2 ip6-allrouters

[schuetzm]

Please reopen, the bugfix in 84fdc1b does not work if the addresses for localhost are ["127.0.0.1", "::1"].

To reproduce, put the following lines into your /etc/hosts:

127.0.0.1   localhost
::1         localhost ipv6-localhost ipv6-loopback

[daya]

@evanphx please reopen this issue @schuetzm steps are enough to reproduce the issue

[schuetzm]

@daya Which error are you getting exactly? "Address already in use", or "Address not available"? If it's the latter, it's likely a misconfiguration on your side, see #1062.

[delner]

Getting this error in 3.6.2, while running in Rails & Docker via rails s:

/usr/local/bundle/gems/puma-3.6.2/lib/puma/binder.rb:266:in `initialize': Cannot assign requested address - bind(2) for "::1" port 3000 (Errno::EADDRNOTAVAIL)
	from /usr/local/bundle/gems/puma-3.6.2/lib/puma/binder.rb:266:in `new'
	from /usr/local/bundle/gems/puma-3.6.2/lib/puma/binder.rb:266:in `add_tcp_listener'

/etc/hosts:

127.0.0.1	localhost
::1	localhost ip6-localhost ip6-loopback
fe00::0	ip6-localnet
ff00::0	ip6-mcastprefix
ff02::1	ip6-allnodes
ff02::2	ip6-allrouters
172.20.0.5	4b1ea4eb06a8

Running with rails s -b 0.0.0.0 allows it to start.

[nateberkopec]

If you have both 127.0.0.0 and ::1 pointing to localhost, puma will try to bind to both and then fail. I'm not sure this is really a bug, it's doing exactly what it's supposed to do. Pointing ::1 to localhost is user error IMO, you should just not do that. LMK if there's some reason why you have to/must do this, otherwise I'm closing.

[delner]

@nateberkopec Just an FYI, that /etc/hosts is not a user specific configuration, thats the default from running Ubuntu in Docker. Anyone who uses that image in Docker (a ton of people) could possibly have the same problem.

[nateberkopec]

running Ubuntu in Docker

Oh alright. I'll reopen until I get a repro. I'm looking at this in #1318.

[delner]

If you'd like to reproduce the hosts file state via Docker, you can do that via docker run --rm ubuntu cat /etc/hosts. As far as Puma itself is concerned, I don't have a repro for you.

[nateberkopec]

Fixed in #1318

[gcaracausi]

You have to modify /etc/hosts and remove localhost from the ipv6 line.
From ::1 localhost ip6-localhost ip6-loopback to ::1 ip6-localhost ip6-loopback

[dimisjim]

@gcaracausi @nateberkopec Indeed, puma binding works when you remove the localhost entry from the ::1 line.

Can easily be done with a dockerfile such as this:

RUN cp /etc/hosts /etc/hosts2 && sed -i 's/::1	localhost/::1/g' /etc/hosts2
CMD cat /tmp/hosts >> /etc/hosts