We spent a whole week rewriting nouveau’s website — the drivers for NVIDIA cards. It started as a one-person effort, but it led to a few people helping me out. We addressed several issues in the nouveau website and improved it a lot. The redesign is live on nouveau.freedesktop.org. In this article, we’ll go over the problems with the old site and the work we’ve done to fix them.
Are you using Linux at work without systemd? Seems unlikely. All our 400+ nodes run RHEL and consequently systemd. This doesn’t seem to impact our researchers’ use of CUDA in the slightest when executing code on the nodes or in any kind of container.
No I’m a big fan of systemd. Unit files for docker deployments is so much easier than anything sysvinit is going to be able to do without a fuckton of logic to write custom configuration files.
Are you using Linux at work without systemd? Seems unlikely. All our 400+ nodes run RHEL and consequently systemd. This doesn’t seem to impact our researchers’ use of CUDA in the slightest when executing code on the nodes or in any kind of container.
No I’m a big fan of systemd. Unit files for docker deployments is so much easier than anything sysvinit is going to be able to do without a fuckton of logic to write custom configuration files.
With other init systems you don’t have to write any custom config files. You just have to start docker; it already has container maintenance built-in.
I’ll never understand why they had to complicate it and require every container to also have a unit of explicit management.
Restart on failure? Start on boot? That’s not automatic.
It is, it’s what
restart: always
does. It will restart a container on failure and start it on boot, unless explicitly stopped.