Hi,

Last Friday Kuryr team noticed elevated rate of failures of the kuryr- kubernetes gate jobs. Later on we've tracked it down to oom-killer slaying our Amphora instances. We couldn't find a reason why that only started to happen on Friday, until now. So commit [1] reduced default swap size to 1 GB. If your jobs need more memory - you can try overwriting that in your job configs like we do in [2].

I'm posting this to save people from a week of debugging as it wasn't a fun activity. ;)

Thanks, Michał

[1] https://opendev.org/openstack/openstack-zuul-jobs/commit/45f555fdf036de786b5... [2] https://review.opendev.org/#/c/752233/4/.zuul.d/base.yaml@35