Really old bugs (semaphore leak) still in Apache???

--00032555fe5ec06b7604746ab45b
Content-Type: text/plain; charset=ISO-8859-1

[note that this is possibly not am od_perl bug]

After a little over a month of using apache, and restarting it many times, I
started getting errors such as this one:

No space left on device: Couldn't create accept lock

The issue is nothing like what the error message suggests: it is not a space
on device issue, the issue is that apache leaks semaphores upon restarts and
runs out of available semaphores.

This is on Ubuntu Hardy

I had the same issue on my old Fedora box, so I quickly adapted an old shell
function to delete all semaphores owned by www-data.

But my question is, Come On! Shouldn't this bug be fixed by now?

As of now, I added semaphore cleaning to my custom "restart apache" script.
This script already addresses a very obnoxious issue that
"/etc/init.d/apache2 stop" does not always stop apache2. My script does
/etc/init.d/apache2 stop, waits 2 seconds and murders every remaining
instance of apache2. Now it also cleans semaphores.

But I got to wonder if this bug is ever going to be fixed?

i

--00032555fe5ec06b7604746ab45b
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable

[note that this is possibly not am od_perl bug]

After a little over =
a month of using apache, and restarting it many times, I started getting er=
rors such as this one:

No space left on device: Couldn't create=
accept lock


The issue is nothing like what the error message suggests: it is not a =
space on device issue, the issue is that apache leaks semaphores upon resta=
rts and runs out of available semaphores.

This is on Ubuntu Hardy r>

I had the same issue on my old Fedora box, so I quickly adapted an old =
shell function to delete all semaphores owned by www-data.

But my q=
uestion is, Come On! Shouldn't this bug be fixed by now?

As of n=
ow, I added semaphore cleaning to my custom "restart apache" scri=
pt. This script already addresses a very obnoxious issue that "/etc/in=
it.d/apache2 stop" does not always stop apache2. My script does /etc/i=
nit.d/apache2 stop, waits 2 seconds and murders every remaining instance of=
apache2. Now it also cleans semaphores.


But I got to wonder if this bug is ever going to be fixed?

i


--00032555fe5ec06b7604746ab45b--

Re: Really old bugs (semaphore leak) still in Apache???

On Fri, Sep 25, 2009 at 10:59 AM, Igor Chudov wrote:
> [note that this is possibly not am od_perl bug]
>
> After a little over a month of using apache, and restarting it many times, I
> started getting errors such as this one:

Is this an apache bug, or a mod_perl bug? If it is an apache bug,
have you tried the httpd list?


>
> No space left on device: Couldn't create accept lock
>
> The issue is nothing like what the error message suggests: it is not a space
> on device issue, the issue is that apache leaks semaphores upon restarts and
> runs out of available semaphores.
>
> This is on Ubuntu Hardy
>
> I had the same issue on my old Fedora box, so I quickly adapted an old shell
> function to delete all semaphores owned by www-data.
>
> But my question is, Come On! Shouldn't this bug be fixed by now?
>
> As of now, I added semaphore cleaning to my custom "restart apache" script.
> This script already addresses a very obnoxious issue that
> "/etc/init.d/apache2 stop" does not always stop apache2. My script does
> /etc/init.d/apache2 stop, waits 2 seconds and murders every remaining
> instance of apache2. Now it also cleans semaphores.
>
> But I got to wonder if this bug is ever going to be fixed?
>
> i
>