sysvinit replacement?
Shane Stixrud
shane at geeklords.org
Sat Oct 15 18:49:18 UTC 2005
On Sat, 15 Oct 2005, Peter Arremann wrote:
>
> The biggest issue with SMF is the state database. A single type on a service
> file can corrupt your state database. So can interrupting a svcs command at
> the wrong time. If that happens, good luck - you go back to the last saved
> state and hope things still work. Of course if you installed software since
> then you end up with a mess and...
Are you saying this is a fundamental problem with a SMF type management
framework or just a implementation detail in the current SMF system?
>
> Then the startup time gains are overrated... Log into your Sol10 box as soon
> as its possible and run "format" - if your box survives that (some older ones
> dont) then you'll see your format stall for a long time before you get your
> output. In the end, if you measure the time from push of the power button
> (assuming a reasonable diag level) to the time you logged in and got your
> format output there is barely any difference between Sol9 and Sol10. Same
> holds true for printing or a JES webserver request.
I never found the startup time issue that big of a deal, it is a bigger
deal for desktops/hand helds than servers IMO. I am more fond of the
service monitoring/event features than I am how quick it is.
>
> Other problems are that you're trading a well known, well debugged system
> against something noone knows and that has virtually no tools to troubleshoot
> at the moment. You also make it much harder for anyone to port software to
> your Linux flavor.
This is true every time major change occurs, it is valid consideration
though.
Cheers,
Shane
More information about the fedora-devel-list
mailing list