is 5.x still too unstable?

[Scott Lambert]

On Wed, Sep 10, 2003 at 04:52:34PM -0400, Tim Middleton wrote:
> 
> I am hoping to move some of the servers in our ISP to FreeBSD. I have been 
> rather hoping 5.2 would be reliable enough, so that we can move to it and 
> enjoy all the -CURRENT goodness. 
> 
> what seems to have taken the test box down yet; but it's been on my mind to 
> solicit opinions here before this happened, so... any thoughts or experiences 
> running 5x on ISP servers to share out there? Are some snapshots known to be 
> better than others? Any tips/tweaks on making 5.x just a little more 
> stable---even at the cost of performance---than a default install (like 
> disabling acpi, as the first thing).

I have my workstation and one not so critical server on 5.x.  While they
tend to run fine most of the time, there is a performance penalty even
with all the debug knobs turned off.  Buildworld will probably always
take longer of 5.x but on my system, I went from a 35 minute buildworld
on 4.x to a 122 minute buildworld on 5.x My kernel builds (including
modules) are now about 31 minutes.

Also, since it is still -CURRENT, the developers tend to still be
ripping out entire subsystems and replacing them with brand new code.

I would say, you don't want to consider going to 5.x on production
servers at least until 5.2 is -RELEASE.  Move your workstations to 5.x
now if you want.

I'll wait a while after 5.2 to see what is happening development-wise
to get a feel for whether or not the codebase has stablized enough to
consider taking the production boxes there.

I am in no rush to abandon 4.x.

-- 
Scott Lambert                    KC5MLE                       Unix SysAdmin
lambert at lambertfam.org