Arthur Pemberton wrote:
On Tue, Mar 25, 2008 at 2:17 PM, Ian Chapman
<packages@xxxxxxxxxxxxxxxxxx> wrote:
Roger Heflin wrote:
>
> The problem is that the enterprise OS's ship buggy kernels too, I have
Of course they do, I'd be surprised to hear if anyone has shipped a 100%
bug free kernel, whether it be Linux, Solaris, AIX, Windows or whatever.
The point though is that RHEL is less of a moving target, if you current
setup works then an update to the kernel is less likely to break
anything, than say Fedora which frequently ships new versions of the
kernel.
One would hope that that these people complaining abut imperfect
kernels in RHEL are find these bugs on their test boxes _before_ they
deploy to production.
Not likely, one of the bugs I found happens with a single application, the test
setup was run it in a job of 100 machines together, and by morning one of the
machines in the job will deadlock, not something that can be found with a single
test box, but it was consistent, the customer could run in on 3 separate sets of
100 overnight, and all 3 would have a single machine that locked up overnight,
reboot the 3 and/or exclude the 3 "bad" machines and try again, and a completely
different 3 machines locked up.
> And the second you add a driver and/or XFS on to RHEL5 you are
> now tainted and *UNSUPPORTED*.
Compared to Fedora where you are *UNSUPPORTED* at the offset?
is XFS even in the vanilla kernel?
Pretty much with RHEL you are unsupported once you install and configure it, at
that point you only really have update support, I have actually called their
support several different times with real actual kernel problems and the support
that was provided only wasted everyones time.
Yes, it is, Sles ships it, and it is in kernel.org kernels. Since RHEL does not
have XFS that means that RHEL is limited to a max 8TB filesystem (this is the
actual ext3 limit on 64bit machines), which eliminated it from usage for several
enterprise customers.
RHEL5 shipped without drivers for Areca and 3ware PCI-e cards even though the
drivers were out for months at the time (the drivers were in vanilla 2.6.19, 1
rev over 2.6.18 used by RHEL5, but Redhat for some reason did not backport the
driver for what is basically an enterprise class raid board for the initial
release).
