Gitlab-backup create using a lot of swap

We have a GitLab CE 14.05 version running in a docker swarm. It has 8 vCPUs, 16 GB of memory and 4 GBs of swap space.

Whenever we run gitlab-backup create it will not complete because it runs out of swap space. The VM will then crater.

I’ve increased the swap space to 20 GB (!) but everyone here is afraid to retry the backup. So we’ve gone 5 weeks without a backup.

20 GB seems totally excessive. Anyone else had to do the same thing?

In my experience, 16Gb RAM should be enough to run a GitLab instance with 1000 active users, and running gitlab-backup create shouldn’t be a RAM hog.

I suspect this may be an issue with how resources are being distributed among the containers in your docker-swarm deployment.