Page MenuHomePhabricator

Questions about Trove guest images
Closed, ResolvedPublic

Description

Two things:

  • The upstream Trove images are built with Ubuntu. Should we just use those, or do we want to build custom images to retain Debian purity
    • Keeping in mind that we are dangerously close to being able to use stock upstream images for our standard cloud-vps VMs, so committing to custom trove builds would mean keeping all of those processes (and bootstrap-vz?) alive.
  • In version Train, Trove guest images use locally-hosted databases. Starting in version V, the guest images wrap the database in docker. That means that starting with V there is only a single base image needed, which makes upgrades and maintenance more straightforward. Is it worth waiting for V so that we can start out doing this the simpler way?

Event Timeline

WMCS has recently been trying to move closer to upstreams as much as possible, including the adoption of tools and practices we might not prefer. I don't think this is an area we have a need or desire to deviate from upstream on. So, I would vote to utilize upstream images for this reason.

Will those VMs be managed via Puppet and other similar WMF-used tools or will Trove manage those VMs themselves without our own systems?

Will those VMs be managed via Puppet and other similar WMF-used tools or will Trove manage those VMs themselves without our own systems?

I don't know yet for certain, but my hope is that they're entirely managed by Trove and we can treat them like black boxes.

Andrew claimed this task.

For now we are using the upstream 'testing' images from https://tarballs.opendev.org/openstack/trove/images/ -- I've discussed this with the devs and it sounds like this will be as secure as building our own.