Page MenuHomePhabricator

tools.topicmatcher update_items_from_sparql.php frequently running Toolforge nodes out of resources
Closed, ResolvedPublic

Description

Hi @Magnus,

update_items_from_sparql.php runs have been frequently eating up all CPU and RAM assigned to the grid exec node it’s running on, which is causing those nodes to go unresponsive until the kernel kills the process and creates all sorts of problems on the grid as a whole. Could you please look if there is anything that can be done to reduce the resource use of this tool to avoid these issues?

In case it's helpful on narrowing down the problematic jobs, most of those have been occurring about 25 minutes past an hour.

Thanks!

Event Timeline

Majavah triaged this task as High priority.May 10 2021, 5:35 PM
Majavah created this task.

This is still happening despite php limits attempted on the servers.

Mentioned in SAL (#wikimedia-cloud) [2021-07-19T16:43:45Z] <bstorm> cleared queue error state caused by excessive resource use by topicmatcher T282474

First time I see this issue here, will investigate.

I think I had a misconfiguration in the cronfile, this is fixed now. Please re-open/ping me if this still happens.