You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@trafficcontrol.apache.org by "Dylan Volz (JIRA)" <ji...@apache.org> on 2017/08/03 16:01:00 UTC

[jira] [Commented] (TC-497) delivery service updates taking >20 secs

    [ https://issues.apache.org/jira/browse/TC-497?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16112994#comment-16112994 ] 

Dylan Volz commented on TC-497:
-------------------------------

https://github.com/apache/incubator-trafficcontrol/pull/763 created

> delivery service updates taking >20 secs
> ----------------------------------------
>
>                 Key: TC-497
>                 URL: https://issues.apache.org/jira/browse/TC-497
>             Project: Traffic Control
>          Issue Type: Bug
>          Components: Traffic Ops
>            Reporter: Dylan Volz
>            Assignee: Dylan Volz
>
> delivery service updates have begun taking about 25 seconds to finish. 
> I have tracked this down to this call:
> my $rs = $self->db->resultset('DeliveryserviceRegex')->search(undef, {prefetch => [
> {regex => undef}
> ,
> {deliveryservice => undef}
> ]} );
> in the find_existing_host_regex helper defined in MojoPlugins/DeliveryService.pm
> The worker is selected for graceful shutdown due to no heartbeat each time an update is made, but usually completes before the shutdown allowing the update to go through.
> If you add a second host regex to the delivery service the update will always fail because it makes that db call twice and thus the worker is killed before the request finishes, and the request is sent to a new worker, continuing the process until the browser cancels the request.
> Note: delivery service updates via the API do not have this problem because it appears that code path does not check for conflicting regexes.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)