The segfault monitor aggressive mode should be optional

Ok now i can reproduce!

i run a query on a data module table (some external table)

  $query = db_select('objecten', 'o');
  $query->addField('o', 'id');
  $query->addField('o', 'item_id');
  $query->condition('gid', $gid, '=');
  $query->condition('query', $query, '=');
  $query->groupBy('query');
  $results = $query->execute();

I get this error:

502 Bad Gateway 

(Even when just running the above code on devel/php)
After that some magic happens and the site is offline. After some time a site verify task seems to run and the site is back

My octopus.log

Sun May  6 14:41:49 CEST 2012 / Debian.squeeze x86_64 / Aegir BOA-2.0.2 / Octopus BOA-2.0.2 / FPM 5.2
Sat May 19 20:10:56 CEST 2012 / Debian.squeeze x86_64 / Aegir BOA-2.0.3 / Octopus BOA-2.0.3 / FPM 5.2 / CLI 5.2

My barracuda.log

Sun May  6 03:06:45 CEST 2012 / Debian.squeeze x86_64 XEN / Aegir BOA-2.0.2 / Barracuda BOA-2.0.2 / Nginx 1.0.12 / PHP 5.2.17 and 5.3.10 / FPM 5.2 / MariaDB localhost / Wildcard NO
Sat May 19 13:50:44 CEST 2012 / Debian.squeeze x86_64 XEN / Aegir BOA-2.0.3 / Barracuda BOA-2.0.3 / Nginx 1.3.0 / PHP 5.2.17 and 5.3.13 / MODERN-NO / FPM 5.2 / CLI 5.2 / MariaDB-5.5.23 localhost / Wildcard NO
Wed May 23 15:57:40 CEST 2012 / Debian.squeeze x86_64 XEN / Aegir BOA-2.0.3 / Barracuda BOA-2.0.3 / Nginx 1.3.0 / PHP 5.2.17 and 5.3.13 / MODERN-NO / FPM 5.2 / CLI 5.2 / MariaDB-5.5.23 localhost / Wildcard NO

.user.octopus.cnf

###
### Configuration created on 120506-1421 with
### Octopus version BOA-2.0.2
###
### NOTE: the group of settings displayed bellow
### will *override* all listed settings in the Octopus script.
###
_USER="user"
_MY_EMAIL="email"
_PLATFORMS_LIST="D7P D7S D7D D6P D6S D6D OAM"
_AUTOPILOT=NO
_HM_ONLY=NO
_O_CONTRIB_UP=YES
_DEBUG_MODE=NO
_MY_OWNIP=
_FORCE_GIT_MIRROR=""
_THIS_DB_HOST=localhost
_DNS_SETUP_TEST=YES
_HOT_SAUCE=NO
_USE_CURRENT=YES
_REMOTE_CACHE_IP=127.0.0.1
_LOCAL_NETWORK_IP=
###
### NOTE: the group of settings displayed bellow will be *overriden*
### by config files stored in the /data/disk/user/log/ directory,
### but only on upgrade.
###
_DOMAIN="somedomain"
_CLIENT_EMAIL="someemail"
_CLIENT_OPTION="CLASSIC"
_CLIENT_SUBSCR="M"
_CLIENT_CORES="4"
###
### Configuration created on 120506-1421 with
### Octopus version BOA-2.0.2
###
_PHP_FPM_VERSION=5.2
_PHP_CLI_VERSION=5.2

aaahh... just wanted to post this!

May 23 23:21:47 some kernel: [1502196.806709] php-fpm[39830]: segfault at 7fffc3dacf78 ip 000000000067549a sp 00007fffc3dacf40 error 6 in php5-fpm[400000+821000]
May 23 23:21:47 felix mysqld: 120523 23:21:47 [Warning] Aborted connection 351669 to db: 'site' user: 'site' host: 'localhost' (Unknown error)
May 23 23:22:01 felix /USR/SBIN/CRON[45041]: (root) CMD (bash /var/xdrago/second.sh >/dev/null 2>&1)
May 23 23:22:01 felix /USR/SBIN/CRON[45040]: (root) CMD (bash /var/xdrago/runner.sh >/dev/null 2>&1)
May 23 23:22:01 felix /USR/SBIN/CRON[45039]: (root) CMD (bash /var/xdrago/minute.sh >/dev/null 2>&1)

I got no mail and i'm the site owner. I also checked the mail.log. But more important what can i do to fix this? Why is a simple query causing a segfault?

Hm... i restarted php-fpm but the segfault is still there. Should i upgrade to php-fpm 5.3?

Sorry it was late yesterday. I had a good look at the links and the php bugs reports. I checked the whole platform for php files with 4096 or 8192 bytes but no results

find ./* -type f -exec ls -la {} \; | grep 4096
find ./* -type f -exec ls -la {} \; | grep 8192

Thanks for all your help! Just for others in 2.0.3 you can disable the segfault error system by renaming or deleting:

/var/xdrago/monitor/check/segfault_alert

Ok in the code i'm doing something extremely stupid:

  $query = db_select('objecten', 'o');
  $query->addField('o', 'id');
  $query->addField('o', 'item_id');
  $query->condition('gid', $gid, '=');
  $query->condition('query', $query, '=');
  $query->groupBy('query');
  $results = $query->execute();

I use a variable $query and the query object is $query. It's mean that php is punishing me for that with a segmentation fault but it's stupid too!