Skip to content
Register

systemd or OpenRC on your VPS?

VPS Hosting by wendy 38 replies 3.8K views
wendy

Member OP

3 #1

In my experience, this debate usually gets heated fast, so I am curious what the actual numbers look like. I run a mix of Windows Server and the occasional Linux box, so I do not have a strong horse in this race. To be fair, systemd gets a lot of flak but it does consolidate a lot of tooling that used to be scattered across a dozen packages. OpenRC boots faster and feels cleaner if you are running Gentoo or Alpine. Then again, I know people who switched to runit just to be different.

What is actually running on your VPS right now, and did you actively choose it or just stick with what the template gave you?

Have you tried restarting it?
quinnbizz

Member

4 #2

Whatever came installed, never changed. I am not about to spend an hour rebuilding init when I could be hunting for the next deal. My current box is a steal at $2.50/GB and the template had systemd. If it had openrc I would run that. If it had sysvinit I would run that. Time is money and I am not made of money

...
garykwh

Member

Inference on a Budget

1 #3

Ngl I actually did switch once, wasted a whole afternoon, broke nvidia drivers, went back to systemd. Now I just grab whatever the cuda toolkit expects. For inference workloads you want the path of least resistance, not some purity test. My 24GB vram box runs systemd and I have never once thought about it during a training run

CUDA cores are my love language
MA

Member

#4

I run whatever the provider ships 😊 never saw a reason to change systemd works fine on my 512MB box openrc probably lighter though I do not benchmark init systems life is too short for that

olya

Member

Always Free

#5

Pro tip: if you are on Oracle Cloud free tier, the ARM images ship systemd and you should probably leave them alone. The automation scripts Oracle provides assume systemd targets.

Heads up: I have seen people break their always-free instances by trying to convert to OpenRC, then the OCI agent stops reporting health and the instance gets reclaimed. Steps to avoid this:

1. Snapshot first. The free tier allows 5 boot volumes, use one.
2. Test your conversion on an ephemeral VM, not your permanent one.
3. If you must switch, disable the OCI agent services properly, do not just mask them.

I have six free tier instances. All systemd. Never touched them. Uptime speaks for itself.

licensing is a suggestion
BL

Member

#6

Time is money and I am not made of money. Same.

LU

Member

#7

Montreal here. I run Alpine on a couple of cheap OVH boxes and systemd on a Hetzner box. OpenRC on Alpine is fine until you need something that only ships a systemd service file, then you are writing your own init scripts. Not fun at 2am when a service dies.

garykwh

Member

Inference on a Budget

#8
olya said:
The automation scripts Oracle provides assume systemd targets

This is the thing people miss. It is not about which init is "better." It is about whose automation you are buying into. I do not care if systemd is bloated. I care that my CUDA stack, my container runtime, and my cloud provider all test against it. That is the real lock-in, not the init system itself.

CUDA cores are my love language
StuSpock

Member

NTP fundamentalist

#9
quinnbizz said:
Time is money and I am not made of money

Same here except I will spend three hours tuning NTP to save 50ms of clock skew. We all have our priorities.

JA

Member

1 #10

I switched a Debian box to OpenRC once because I was running it on a 256MB VPS. It worked, booted faster, but then I needed to run Docker and the official install assumes systemd. Spent more time fighting workarounds than I saved in RAM. Back to systemd, box has 512MB now, problem solved with money not config.

Post a reply

You need an account to reply. Log in or register to join the conversation.

Post reply Preview Save draft