php7.3-fpm segfaults

classic Classic list List threaded Threaded
10 messages Options
Reply | Threaded
Open this post in threaded view
|

php7.3-fpm segfaults

Lucio Crusca-2
Hello,

I'm running buster/sid on my notebook and on a desktop pc. I'd like to
use both for PHP software development, so I've setup Apache2 and
php7.3-fpm to create a devel env similar to the real production servers.

On the destop pc it works. On the notebook it segfaults. Please note
that the two systems are not exactly the same, with the notebook being a
older debian install, upgraded several releases since the first install.

Here is what i get in the logs:

# cat /var/log/php7.3-fpm.log | tail -n 7
[12-Jan-2019 08:57:36] NOTICE: fpm is running, pid 815
[12-Jan-2019 08:57:36] NOTICE: ready to handle connections
[12-Jan-2019 08:57:36] NOTICE: systemd monitor interval set to 10000ms
[12-Jan-2019 08:59:45] WARNING: [pool lucio] child 1635 exited on signal
11 (SIGSEGV) after 129.612030 seconds from start
[12-Jan-2019 08:59:45] NOTICE: [pool lucio] child 3635 started
[12-Jan-2019 09:09:44] WARNING: [pool lucio] child 3635 exited on signal
11 (SIGSEGV) after 598.434794 seconds from start
[12-Jan-2019 09:09:44] NOTICE: [pool lucio] child 23618 started

The last 2 lines repeat each time I load a page in the browser. The page
obviously displays only a "Service Unavailable" message from Apache.

How do I get more details about what's going on with PHP-FPM?



Reply | Threaded
Open this post in threaded view
|

Re: php7.3-fpm segfaults

Lucio Crusca-2
Il 12/01/19 10:25, Lucio ha scritto:

>
> How do I get more details about what's going on with PHP-FPM?
>

My question holds, but I have discovered something new: if I purge the
php7.3-mysql package, php7.3-fpm starts working again, and I get:

"Your PHP installation appears to be missing the MySQL extension which
is required by WordPress."

which shows php7.3-fpm is working. When I reinstall php7.3-mysql, it
starts crashing again.


Reply | Threaded
Open this post in threaded view
|

Re: php7.3-fpm segfaults

Roberto C. Sánchez-2
In reply to this post by Lucio Crusca-2
On Sat, Jan 12, 2019 at 10:25:51AM +0100, Lucio wrote:

> Hello,
>
> I'm running buster/sid on my notebook and on a desktop pc. I'd like to use
> both for PHP software development, so I've setup Apache2 and php7.3-fpm to
> create a devel env similar to the real production servers.
>
> On the destop pc it works. On the notebook it segfaults. Please note that
> the two systems are not exactly the same, with the notebook being a older
> debian install, upgraded several releases since the first install.
>
> Here is what i get in the logs:
>
> # cat /var/log/php7.3-fpm.log | tail -n 7
> [12-Jan-2019 08:57:36] NOTICE: fpm is running, pid 815
> [12-Jan-2019 08:57:36] NOTICE: ready to handle connections
> [12-Jan-2019 08:57:36] NOTICE: systemd monitor interval set to 10000ms
> [12-Jan-2019 08:59:45] WARNING: [pool lucio] child 1635 exited on signal 11
> (SIGSEGV) after 129.612030 seconds from start
> [12-Jan-2019 08:59:45] NOTICE: [pool lucio] child 3635 started
> [12-Jan-2019 09:09:44] WARNING: [pool lucio] child 3635 exited on signal 11
> (SIGSEGV) after 598.434794 seconds from start
> [12-Jan-2019 09:09:44] NOTICE: [pool lucio] child 23618 started
>
> The last 2 lines repeat each time I load a page in the browser. The page
> obviously displays only a "Service Unavailable" message from Apache.
>
> How do I get more details about what's going on with PHP-FPM?
>

Are there corresponding entries in the Apache log?  Can you turn the
Apache log level up to debug or higher and cause the failure again?

Regards,

-Roberto
--
Roberto C. Sánchez

Reply | Threaded
Open this post in threaded view
|

Re: php7.3-fpm segfaults

Lucio Crusca-2
Il 13/01/19 01:36, Roberto C. Sánchez ha scritto:
> Are there corresponding entries in the Apache log?

Not much to care about actually, because Apache is configured as reverse
proxy for php-fpm. Even raising log level I only get details about what
Apache is doing, not about what PHP is doing.

However here is the Apache error log at debug level, just in case I'm wrong:

https://t2m.io/82KHik8J

Reply | Threaded
Open this post in threaded view
|

Re: php7.3-fpm segfaults

Roberto C. Sánchez-2
On Sun, Jan 13, 2019 at 01:31:07PM +0100, Lucio wrote:

> Il 13/01/19 01:36, Roberto C. Sánchez ha scritto:
> > Are there corresponding entries in the Apache log?
>
> Not much to care about actually, because Apache is configured as reverse
> proxy for php-fpm. Even raising log level I only get details about what
> Apache is doing, not about what PHP is doing.
>
> However here is the Apache error log at debug level, just in case I'm wrong:
>
> https://t2m.io/82KHik8J
>

I agree that there is nothing much interesting there.

Regards,

-Roberto
--
Roberto C. Sánchez

Reply | Threaded
Open this post in threaded view
|

Re: php7.3-fpm segfaults

Roberto C. Sánchez-2
On Sun, Jan 13, 2019 at 09:13:24AM -0500, Roberto C. Sánchez wrote:

> On Sun, Jan 13, 2019 at 01:31:07PM +0100, Lucio wrote:
> > Il 13/01/19 01:36, Roberto C. Sánchez ha scritto:
> > > Are there corresponding entries in the Apache log?
> >
> > Not much to care about actually, because Apache is configured as reverse
> > proxy for php-fpm. Even raising log level I only get details about what
> > Apache is doing, not about what PHP is doing.
> >
> > However here is the Apache error log at debug level, just in case I'm wrong:
> >
> > https://t2m.io/82KHik8J
> >
>
> I agree that there is nothing much interesting there.
>
I forgot to add that the persistent segfault looks like it could be
either a defect in php-fpm or a hardware problem (faulty memory).  Given
the absence of similar reports (which would indicate software as the
more likely culprit), I would lean toward hardware.  Have you considered
running memtest on your system?

Regards,

-Roberto

--
Roberto C. Sánchez

Reply | Threaded
Open this post in threaded view
|

Re: php7.3-fpm segfaults

Lucio Crusca-2
Il 13/01/19 15:37, Roberto C. Sánchez ha scritto:
> I would lean toward hardware.  Have you considered
> running memtest on your system?

The problem is perfectly reproducible on my system.
Removing php7.3-mysql the problem goes away. When I install it again I
get the problem back too. I hardly see that being caused by a faulty RAM
module. Besides in my experience the first consequence of faulty RAM on
Linux is a filesystem corruption.

Anyway I've double checked and ran memtest86+. No errors reported.

I suspect the problem could be some old library files lingering around,
because this system was first installed several years ago and tracked
testing/sid throughout all those years. That would explain why there
aren't any reports out there of similar problems.

However I don't know how to verify my hypothesis.

Reply | Threaded
Open this post in threaded view
|

Re: php7.3-fpm segfaults

Georgi Naplatanov
On 1/13/19 7:17 PM, Lucio wrote:
> Il 13/01/19 15:37, Roberto C. S�nchez ha scritto:
> I suspect the problem could be some old library files lingering around,
> because this system was first installed several years ago and tracked
> testing/sid throughout all those years. That would explain why there
> aren't any reports out there of similar problems.
>

Hi Lucio,

you can uninstall absolute packages with aptitude.

Don't worry about your old initial installation, I did mine 8 years ago
and I've never had any problem.

HTH

Kind regards
Georgi

Reply | Threaded
Open this post in threaded view
|

Re: php7.3-fpm segfaults

Curt
In reply to this post by Roberto C. Sánchez-2
On 2019-01-13, Roberto C  Sánchez <[hidden email]> wrote:

> On Sun, Jan 13, 2019 at 09:13:24AM -0500, Roberto C. Sánchez wrote:
>> On Sun, Jan 13, 2019 at 01:31:07PM +0100, Lucio wrote:
>> > Il 13/01/19 01:36, Roberto C. Sánchez ha scritto:
>> > > Are there corresponding entries in the Apache log?
>> >
>> > Not much to care about actually, because Apache is configured as reverse
>> > proxy for php-fpm. Even raising log level I only get details about what
>> > Apache is doing, not about what PHP is doing.
>> >
>> > However here is the Apache error log at debug level, just in case I'm wrong:
>> >
>> > https://t2m.io/82KHik8J
>> >
>>
>> I agree that there is nothing much interesting there.
>>
> I forgot to add that the persistent segfault looks like it could be
> either a defect in php-fpm or a hardware problem (faulty memory).  Given
> the absence of similar reports (which would indicate software as the
> more likely culprit), I would lean toward hardware.  Have you considered
> running memtest on your system?

Is this related?

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=913635

(where it's suggested to install php7.3-fpm-dbgsym)

If not, mille excuses.

> Regards,
>
> -Roberto
>


--
He used sentences differently from any other prose writer. He always sounded
like a slightly drunk man who is very melancholy, who has no illusions about
life, who is very strong but whose strength is entirely unnecessary.
--Krasznahorkai on Krúdy

Reply | Threaded
Open this post in threaded view
|

Re: php7.3-fpm segfaults

Lucio Crusca-2
Il 15/01/19 13:43, Curt ha scritto:
> Is this related?
>
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=913635
>

Might be, thanks for the pointer. BTW, I've just realized it segfaults
on both systems of mine, not just the notebook. Moreover it does not
segfault immediately, but it actually runs some of the WP code and some
of my WP plugin code, but I've not managed to bisect the exact line yet.
There must be some code path in my WP plugin that leads to the segfault.

Maybe my code has something in common with the Cacti poller, so that bug
applies to me too.

I'll post here (and/or in the bug report) as soon as I have new findings.