On Thu, 09 Oct 2008 17:40:03 +1300
Jasper Bryant-Greene <[EMAIL PROTECTED]> wrote:

> On 9/10/2008, at 4:30 PM, Steve Holdoway wrote:
> 
> > I've got debian servers with uptimes measured in years ( well,  
> > except for the single reboot when they moved data centres about a  
> > year ago ), and I've got CentOS servers in the same category.
> 
> If their uptime is measured in years then I'd bet good money that  
> they're vulnerable to at least one local user root exploit, possibly  
> more. Upgrading kernels isn't just for fun, there's some profit to be  
> had too.
> 
> -jasper

The word local is the important bit here - apart from the fact that most of 
these are in New Jersey. They're servers, serving content to the internet 
through strictly controlled services. These services have been upgraded many 
times - and now run orders of magnitude faster and infinitely more reliably. To 
get access to these servers would be extremely difficult - internet facing 
access is as squeaky new as practicable, and ssh access is extremely strictly 
controlled.

As I said elsewhere, it's all down to risk, and your perception of it. In these 
cases, there are no local users, and internet-facing services ( on the servers 
directly connected to the internet ) are mainly non-standard - and all servers 
are heavily secured, including IDS - so the prime directive for the kernel on 
these servers is to be stable, as far as I'm concerned. 

Yes, you're absolutely correct, there are plenty of reasons to upgrade a 
kernel, but there are also situations where the risk outweighs them. In this 
case, I'll need to start replacing hardware in a year or so, and that's when 
I'll uprade. I must admit, 64bit database servers would be nice (:

Cheers,

Steve

Steve
-- 
Steve Holdoway <[EMAIL PROTECTED]>

Reply via email to