On Thu, Nov 17, 2005 at 11:04:10AM +0100, Morten Nilsen wrote:
> Alain Fauconnet wrote:
> > Maybe a bit exagerated, right, but running a 2.6 kernel on top of TSL
> > 2.2 isn't precisely a smooth operation. It requires modifying
> > /etc/rc.sysinit and installing a foreign mkinitrd RPM (unless you
> > compile all the drivers you need to boot static).
> 
> rpm?

Yes, you know, that file archive format used to package software by
several Linux distributions ;-)
Not sure I understand your question here, but as I mentioned in a
previous posting, I installed a mkinitrd RPM from Fedora Core 3
instead of the TSL 2.2 one to get support for .ko kernel modules.
That was part of the "hacking". OK, my sentence was poor English, it's
TSL 2.2 that has been hacked, not the kernel (although I had to
patch it a bit too, to avoid the '/proc' permission error due to syslog
running non-root on TSL 2.2).

> 
> > So from a production perspective, it's kind of a hack IMHO.
> > I wouldn't call it like this if it were e.g. a contrib RPM with
> > a 2.6 kernel, but I hardly see how this could be done.
> 
> I believe the wiki recommends compiling the stuff you need in (that's 
> what I wrote there way back when, in any case)
> 
> no sense having the stuff you need as modules if your kernel is compiled 
> specifically to your system..

While I would generally agree with this, there are many reasons
that might drive you otherwise:
- when you have more than one SCSI controller and the one that
connects the boot disk isn't detected first due to PCI IDs that you
may not be able to change (e.g. on-board controllers)
- when you have a set of slightly different servers in a farm and you
want to have a common, standardized .config to ease your work burden
etc.

Now to reply to "why not TSL 3.0?":

- these servers are used to provide mail service to a community with
high availability expectations and not much tolerance to downtime: TSL
3.0 isn't considered "production stable" at this point
- I'm quite familiar to TSL 2.2. TSL 3.0 introduces a lot of
new stuff (e.g. the 4-letter acronym boot thing - can't recall the
name now) that looks very hostile to me at this point. I just don't
have time to get acquainted to 3.0 now.
- we use a lot of 3rd party software and homebrewn code too. Some of
it may break on top of 3.0. It will take time until we can validate
all this, probably not until 3.2 is out and considered a stable
release!
- we have a server farm and 2.2 is our standard o/s now. Migrating to
3.x, when time comes, will have to take place on all servers to
preserve consistency. Quite some work and not something we can afford
now.

None of this should sound unusual to people administering
production server farms, I think.

Greets,
_Alain_


_______________________________________________
tsl-discuss mailing list
[email protected]
http://lists.trustix.org/mailman/listinfo/tsl-discuss

Reply via email to