(fixed-ish): Ubuntu 10.10 - usplash bug again?

[Fullmetal Chocobo]

I've been having a weird error on all my LAMP setups (VM and bare-metal), where on restarting the system, it freezes at the splash screen where it asks for the pass-phrase for my Apache SSL key. The only fix I've seen thus far is to use a non-secure key for the SSL (no pass phrase in it). Then the system works fine. The weirdest thing is that it seems to replicate bugs in previous version of Ubuntu:

Bug #139363
Bug #56319
Bug #55159

I also did a test, and installed Ubuntu Server 10.04, and did all of the updates, and now it is happening there as well. On 10.10, I get an error about swap-1 is not available, press s to wait or h to skip (or something like that).

Now I can put in an insecure key, reboot the system, swap key files in the apache config file, then restart apache, and it works. But that seems a bit ridiculous to have to do every time.

Anyone else had anything like this?

 

[Fullmetal Chocobo]

Edit for clarification... (17DEC2010 @ 2226)
I'm unable to start apache2 because the TTY1 is locked with it waiting for the passphrase to my ssl key. I have had this happen on multiple machines now, all with the same problem, on different version of Ubuntu Server and different platforms (both VM and bare-metal).

 

[Nothinman]

Can't you just remove usplash?

 

[KillerBee]

I installed a fresh 10.10 64-bit Server - using VMware Work7
Didn't apply any updates yet and getting same problem

It starts Apache and asks for password by displaying:
Enter pass phrase:

but before you can enter a password
the error message about swap_1 is displayed immediately to the right of the Apache password prompt so they run together

Enter pass phrase:The disk drive for /dev/mapper/ubuntu10-swap_1 is not ready yet or not present
Continue to wait; or Press S to skip mounting or M for manual recovery

ie: The two competing messages are waiting for input but neither is accepting any input

Edit: The last line in /var/log/messages before the reboot shows the swap did finally mount
so would say they aren't both waiting for for input - only the Apache is waiting for input
but since the Apache password-prompt was interrupted by the swap error message - it can't regain control of the screen

 

[KillerBee]

couple other things in case anyone cares:
Using a full encrypted file system with password on startup (password prompt works)
There is no /sbin/usplash - apparently using plymouth instead

Also tried adding a 30 second delay to beginning of Apache startup:
so now the swap error message is displayed before the Apache password-prompt starts - but still can't type in the Apache password when asked
whatever is responsible for displaying the swap error message doesn't release control of screen back to user

Not sure why getting the swap error in the first place but guess it has something to do with the encrypted disk?


Edit2: This older fix mentioned here worked (adding stty sane to the apache startup script)
http://ubuntuforums.org/archive/index.php/t-774692.html

sed -e '/^ENV=/i stty sane' /etc/init.d/apache2|sudo tee /etc/init.d/apache2

 

[Fullmetal Chocobo]

I will try that fix when I'm able to remote to my servers again, or get a chance to get to the office.

Thanks.

 

[Fullmetal Chocobo]

Well, I applied the "fix", and now I get no return from apache2 when I do restarts, etc. I'm limited to CLI right now, so I cannot verify if it worked or not.

 

[KillerBee]

What does "ps -ef |grep apache2" show?

Do you get prompted to enter the password?

Does the script work at all?
sudo /etc/init.d/apache2 status


If not it sounds like the script may be corrupted:
Hopefully you saved the original 🙂


Here's how the /etc/init.d/apache2 change looks on mine (around line 34)

Before change:
==========================================
export APACHE_CONFDIR APACHE_ENVVARS

ENV="env -i LANG=C PATH=/usr/local/bin:/usr/bin:/bin"
==========================================


After change:
==========================================
export APACHE_CONFDIR APACHE_ENVVARS

stty sane
ENV="env -i LANG=C PATH=/usr/local/bin:/usr/bin:/bin"
==========================================

 

[Fullmetal Chocobo]

Below is the results of the commands you stated.

[B]root@x:/# ps -ef |grep apache2[/B]
root      1052   751  0 17:04 ?        00:00:00 /bin/sh -e /etc/rc2.d/S91apache2 start
root      1061  1052  0 17:04 ?        00:00:00 /bin/sh /usr/sbin/apache2ctl start
root      1066  1061  0 17:04 ?        00:00:00 /usr/sbin/apache2 -k start
root      1179  1161  0 17:04 pts/1    00:00:00 grep --color=auto apache2

[B]root@x:/# /etc/init.d/apache2 status[/B]
root@x:/# 

[B]root@bx:/# /etc/init.d/apache2 restart[/B]
 * Restarting web server apache2
(98)Address already in use: make_sock: could not bind to address 0.0.0.0:80
no listening sockets available, shutting down
Unable to open logs
   ...fail!

The 'fix' adds 'stty sane' to /etc/init.d/apache2 (puts it in line 14). There is no change in the operation or booting of the server, and still has the same issues as before.

It is failing to restart because it is still trying to get the passphrase for the SSL key in TTY1.

I'm getting the script from another server and starting over (the servers are configured very similarly).

 

[KillerBee]

line 14? ... it does sound screwed up and trying again is a good idea

If I hold mine at the password prompt
I get same response as yours for these two:
ps -ef |grep apache2
/etc/init.d/apache2 restart

but for "/etc/init.d/apache status" it should return something
mine returns: "Apache2 is NOT running"

To stop it remotely (or when at the machine start a new console)
will need to use "sudo killall apache2" first

ie:Running /etc/init.d/apache2/ "stop" or "restart" won't work
since it hasn't written a pid file yet when waiting for the password

Then should be able to run "/etc/init.d/apache2 start"

 

[Fullmetal Chocobo]

Okay. Redid the server, and added the fix. I can now kill apache2, and restart it. And it also gives a status when I request it from apache2. So apparently there was an issue / conflict with the apache2 script, or it was corrupted.

Pain in the ass, but it is workable now. Thanks for your help. I will be paying attention to the status of this bug...

Only one difference I've noticed between this issue and bug #582963, is that I can't even type in the password in the local console. TTY1 is a blank screen until I kill all apache2 processes. Immediately after that, TTY1 is released and goes to the login prompt.

 

[KillerBee]

Not sure who to follow 😛

SSL pass phrase dialog can't read input
https://bugs.launchpad.net/ubuntu/+source/apache2/+bug/582963
(this adds the apache2 stty sane)

init script resets isig flag in an incorrect manner
https://bugs.launchpad.net/ubuntu/+source/plymouth/+bug/626723
(this comments it back out)

Edit: just to be clear ..
Are you saying even with stty sane in your apache2 script
you still can't enter the password at bootup - without killing apache first?

note: I do have to hit "Enter" once - before typing the password

 

[Fullmetal Chocobo]

Not sure who to follow 😛

SSL pass phrase dialog can't read input
https://bugs.launchpad.net/ubuntu/+source/apache2/+bug/582963
(this adds the apache2 stty sane)

init script resets isig flag in an incorrect manner
https://bugs.launchpad.net/ubuntu/+source/plymouth/+bug/626723
(this comments it back out)

Edit: just to be clear ..
Are you saying even with stty sane in your apache2 script
you still can't enter the password at bootup - without killing apache first?

note: I do have to hit "Enter" once - before typing the password

Okay. I had forgotten about the 'enter'. I can enter the encryption password (it's a laptop, so everything is encrypted just in case). I can't believe this issue exists and it hasn't been fixed yet. Take the good with the bad I guess.

Thank you very much for your help -- this has been incredibly frustrating, but seems to be working out now, albeit roughly.