Page MenuHomePhabricator

Investigate why orphan-slayer job is running for hours
Closed, ResolvedPublic

Description

When looking at the process list on civi1001 we noticed that the orphan-slater for paypal and ingenico was running for 12 and 5 hours plus respectively.

jgleeson@civi1001:/var/spool/prometheus$ ps aux --sort=-%mem,-%cpu | head -n 50
USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
www-data 14647  3.9  4.2 1047620 689480 ?      S    16:00  12:52 php /usr/share/php/drush/drush.php --user=1 -v -r /srv/org.wikimedia.civicrm/drupal orphan-slayer paypal_ec
www-data 26968  4.0  1.5 615740 254836 ?       S    19:00   6:01 php /usr/share/php/drush/drush.php --user=1 -v -r /srv/org.wikimedia.civicrm/drupal orphan-slayer ingenico

Event Timeline

@jgleeson I found a couple of candidates for ways to speed it up:

T249276 (processing messages that we know are failed while donor is on payments-wiki)

T205566 (making extra API calls which are guaranteed to fail after finding out that the payment is unprocessable)

Change 749586 had a related patch set uploaded (by Ejegg; author: Ejegg):

[wikimedia/fundraising/crm@master] Add batch and time_limit options to orphan slayer

https://gerrit.wikimedia.org/r/749586

Change 749586 merged by jenkins-bot:

[wikimedia/fundraising/crm@master] Add batch and time_limit options to orphan slayer

https://gerrit.wikimedia.org/r/749586