[systemd-devel] systemd-udevd and services with DefaultDependencies=no

Thomas HUMMEL thomas.hummel at pasteur.fr
Thu Nov 28 19:03:17 UTC 2024


On 11/28/24 7:56 PM, Cristian Rodríguez wrote:
> On Thu, Nov 28, 2024 at 3:50 PM Thomas HUMMEL <thomas.hummel at pasteur.fr>
> wrote:
> 
>>
>>
>> Hello,
>>
>> coming back to the topic (sorry, I had an actual physical issue with the
>> device which blurred my analysis):
>>
>> 1. On a system where /dev/nvme0, /dev/nvme0n1 and /dev/nvme0n1p1 files
>> exist, systemctl list-units --type=device | grep -i nvme only list
>> physical path ones such as
>> sys-devices-pci0000:a0-0000:a0:03.1-0000:a1:00.0-nvme-nvme0-nvme0n1.device
>> for instance.
>>
>> Did you assume something I do not have by evoking dev-nvme.device ?
>>
>>
>>
> Your assumptions start early.. do the kernel swear to $deity that
> /dev/nvme0 will not change meaning for as long it is the same hardware,
> plugged in the same slot etc.etc.? I don't believe this warranty exists.
> you need to rely on something reasonably stable to begin with. I'm afraid
> this is the wrong premise to start with.
> 

Hello, thanks for your reply.

I know about predictable vs unpredictable names and I agree with you. 
This is the reason why I'm using predictable names for nics for instance.

I'm in a particular context though were hosts are "stateless" (OS is PXE 
booted and lies only in RAM) and each node has exactly and only one nvme 
drive.

Granted I could and probably should reason the other way around and 
assume for the same hardware, drive would be at the same "physical 
location".

Anyway I wonder if it is expected that I don't see any systemd dev-nvme 
device unit.

Thanks for your help

-- 
Thomas HUMMEL
HPC Group
Institut PASTEUR
Paris, FRANCE


More information about the systemd-devel mailing list