« Return to Thread: Seg-faults on highly-trafficed APC site
Re: Seg-faults on highly-trafficed APC site
by KannanM@gmail.com (Kannan Muthukkaruppan)
::
Rate this Message:
We ran into something similar a while back. We were using 5.2.5 at that
time. See this thread:
<
http://markmail.org/message/bhy2kbtwrljecnna#query:kannan%20muthukkaruppan%20apc+page:1+mid:ajsctublcbbjvnax+state:results
>
This issue-- the issue with improper ref count handling in
zend_assign_to_variable when the left-hand side is a reference (is_ref
set)--- was fixed in a later version of PHP. Don't recall the exact version
though.
regards,
Kannan Muthukkaruppan
On Sun, Dec 7, 2008 at 2:32 PM, SinistralEnigma <sean@...> wrote:
>
> Hi,
> I had a decent number of segfaults that appeared to be caused by APC (I am
> running APC v3.0.19 and PHP v5.1.6). Rasmus recommended that I bump up the
> space allotted to APC for caching and that made the problem go away for
> about a month but it recently came back and caused my servers to crash
> again.
>
> I have APC running on two Apache servers (behind a squid caching server)
> and
> I watched apc_INFO.php closely for the whole month, checking several times
> a
> day. Neither of the servers were using even 50% of the space allotted to
> them, and they both had a "Cache full count" of 0 the last time I checked
> (which would have been only a matter of hours before the error occurred).
>
> The only anomaly I saw ahead of time is that the one of the servers (where
> I
> hadn't restarted httpd during the whole 4 weeks) started to have values on
> the apc_INFO panel that were negative. For example, the hits in the
> hits/misses were a large negative number - around 1.9 billion if I recall.
> The "Request Rate" was also negative.
>
> If it matters, this is being used for a MediaWiki v1.13.2 site.
>
> Any hints would be appreciated. I've attached a good deal of what I feel
> is
> probably helpful debugging information below.
> Thanks for your time,
> - Sean Colombo [LyricWiki.org]
>
>
>
>
> 01260000-01268000 rw-p 01260000 00:00 0
> 01268000-013f9000 r-xp 00000000 08:06 52857819
> /usr/lib/php/modules/mbstring.so
> 013f9000-0140c000 rw-p 00191000 08:06 52857819
> /usr/lib/php/modules/mbstring.so
> 02a7d000-02afa000 r-xp 00000000 08:06 52670281
> /usr/lib/libfreetype.so.6.3.10
> 02afa000-02afd000 rw-p 0007d000 08:06 52670281
> /usr/lib/libfreetype.so.6.3.10
>....
>
> /usr/lib/mysql/libmysqlclient.so.15.0.0
> 0769d000-076df000 rw-p 00122000 08:06 52954267
> /usr/lib/mysql/libmysqlclient.so.15.0.0
> 076df000-076e0000 rw-p 076df000 00:00 0
> 09892000-09ed5000 rw-p 09892000 00:00 0
> afe63000-afe77000 rw-s 00000000 00:08 888627839 /dev/zero (deleted)
> afe77000-afef5000 rw-s 00000000 00:08 1021609716 /dev/zero (deleted)
> afef5000-b7ef5000 rw-s 00000000 08:06 29949955 /tmp/apc.RumXNU (deleted)
> b7ef5000-b7f0e000 rw-s 00000000 00:08 1021609714 /dev/zero (deleted)
> b7f0e000-b7f16000 rw-p b7f0e000 00:00 0
> bfd91000-bfda7000 rw-p bfd91000 00:00 0 [stack]
> [Wed Dec 03 11:27:48 2008] [notice] child pid 14537 exit signal Aborted (6)
> *** glibc detected *** /usr/sbin/httpd: free(): invalid pointer: 0x09d4daf4
> ***
> [Wed Dec 03 11:27:54 2008] [notice] child pid 14543 exit signal
> Segmentation
> fault (11)
> [Wed Dec 03 11:27:55 2008] [notice] child pid 14542 exit signal
> Segmentation
> fault (11)
>
> --
> View this message in context:
> http://www.nabble.com/Seg-faults-on-highly-trafficed-APC-site-tp20886787p20886787.html
> Sent from the Pecl - Dev mailing list archive at Nabble.com.
>
>
> --
> PECL development discussion Mailing List (http://pecl.php.net/)
> To unsubscribe, visit: http://www.php.net/unsub.php
>
>
time. See this thread:
<
http://markmail.org/message/bhy2kbtwrljecnna#query:kannan%20muthukkaruppan%20apc+page:1+mid:ajsctublcbbjvnax+state:results
>
This issue-- the issue with improper ref count handling in
zend_assign_to_variable when the left-hand side is a reference (is_ref
set)--- was fixed in a later version of PHP. Don't recall the exact version
though.
regards,
Kannan Muthukkaruppan
On Sun, Dec 7, 2008 at 2:32 PM, SinistralEnigma <sean@...> wrote:
>
> Hi,
> I had a decent number of segfaults that appeared to be caused by APC (I am
> running APC v3.0.19 and PHP v5.1.6). Rasmus recommended that I bump up the
> space allotted to APC for caching and that made the problem go away for
> about a month but it recently came back and caused my servers to crash
> again.
>
> I have APC running on two Apache servers (behind a squid caching server)
> and
> I watched apc_INFO.php closely for the whole month, checking several times
> a
> day. Neither of the servers were using even 50% of the space allotted to
> them, and they both had a "Cache full count" of 0 the last time I checked
> (which would have been only a matter of hours before the error occurred).
>
> The only anomaly I saw ahead of time is that the one of the servers (where
> I
> hadn't restarted httpd during the whole 4 weeks) started to have values on
> the apc_INFO panel that were negative. For example, the hits in the
> hits/misses were a large negative number - around 1.9 billion if I recall.
> The "Request Rate" was also negative.
>
> If it matters, this is being used for a MediaWiki v1.13.2 site.
>
> Any hints would be appreciated. I've attached a good deal of what I feel
> is
> probably helpful debugging information below.
> Thanks for your time,
> - Sean Colombo [LyricWiki.org]
>
>
>
>
> 01260000-01268000 rw-p 01260000 00:00 0
> 01268000-013f9000 r-xp 00000000 08:06 52857819
> /usr/lib/php/modules/mbstring.so
> 013f9000-0140c000 rw-p 00191000 08:06 52857819
> /usr/lib/php/modules/mbstring.so
> 02a7d000-02afa000 r-xp 00000000 08:06 52670281
> /usr/lib/libfreetype.so.6.3.10
> 02afa000-02afd000 rw-p 0007d000 08:06 52670281
> /usr/lib/libfreetype.so.6.3.10
>
>
> /usr/lib/mysql/libmysqlclient.so.15.0.0
> 0769d000-076df000 rw-p 00122000 08:06 52954267
> /usr/lib/mysql/libmysqlclient.so.15.0.0
> 076df000-076e0000 rw-p 076df000 00:00 0
> 09892000-09ed5000 rw-p 09892000 00:00 0
> afe63000-afe77000 rw-s 00000000 00:08 888627839 /dev/zero (deleted)
> afe77000-afef5000 rw-s 00000000 00:08 1021609716 /dev/zero (deleted)
> afef5000-b7ef5000 rw-s 00000000 08:06 29949955 /tmp/apc.RumXNU (deleted)
> b7ef5000-b7f0e000 rw-s 00000000 00:08 1021609714 /dev/zero (deleted)
> b7f0e000-b7f16000 rw-p b7f0e000 00:00 0
> bfd91000-bfda7000 rw-p bfd91000 00:00 0 [stack]
> [Wed Dec 03 11:27:48 2008] [notice] child pid 14537 exit signal Aborted (6)
> *** glibc detected *** /usr/sbin/httpd: free(): invalid pointer: 0x09d4daf4
> ***
> [Wed Dec 03 11:27:54 2008] [notice] child pid 14543 exit signal
> Segmentation
> fault (11)
> [Wed Dec 03 11:27:55 2008] [notice] child pid 14542 exit signal
> Segmentation
> fault (11)
>
> --
> View this message in context:
> http://www.nabble.com/Seg-faults-on-highly-trafficed-APC-site-tp20886787p20886787.html
> Sent from the Pecl - Dev mailing list archive at Nabble.com.
>
>
> --
> PECL development discussion Mailing List (http://pecl.php.net/)
> To unsubscribe, visit: http://www.php.net/unsub.php
>
>
« Return to Thread: Seg-faults on highly-trafficed APC site
| Free embeddable forum powered by Nabble | Forum Help |
