| 1 |
On Mar 18, 2012 3:52 PM, "Canek Peláez Valdés" <caneko@×××××.com> wrote: |
| 2 |
> |
| 3 |
> If the config file doesn't exists, the service will not start, and you |
| 4 |
> can check the reason why with |
| 5 |
> |
| 6 |
> systemctl status sshd.service |
| 7 |
> |
| 8 |
> And of course you can set another mini sevice unit file to create the |
| 9 |
> hostkeys. But I repeat: I think those tasks belong into the package |
| 10 |
> manager, no the init script. |
| 11 |
> |
| 12 |
|
| 13 |
Between installation by package manager and actual execution by the init |
| 14 |
system, things might happen on the required file(s). Gentoo's initscript |
| 15 |
guards against this possibility *plus* providing helpful error messages in |
| 16 |
/var/rc.log |
| 17 |
|
| 18 |
Or, said configuration files might be corrupted; the OpenRC initscript -- |
| 19 |
if written defensively -- will be able to detect that and (perhaps) |
| 20 |
fallback to something sane. systemd can't do that, short of putting all |
| 21 |
required intelligence into a script which it executes on boot. |
| 22 |
|
| 23 |
Now, if one has to put all the intelligence into a script file which gets |
| 24 |
executed by systemd, that results in a system that's more complex than |
| 25 |
plain OpenRC. Not only would one need to maintain the starting script, but |
| 26 |
one must also maintain systemd + dbus. |
| 27 |
|
| 28 |
So, the *only* benefit I can see about systemd is the smarter parallel |
| 29 |
startup of services. And believe me if I say that server guys (the ones |
| 30 |
handling tens or even hundreds of servers) would much prefer sequential |
| 31 |
startup of services than parallel ones. The former is deterministic, the |
| 32 |
latter is not. |
| 33 |
|
| 34 |
Rgds, |
Archives