Another slip in the FC6 schedule
Alfredo Ferrari
list at pceet030.cern.ch
Tue Oct 17 22:46:02 UTC 2006
... indeed we have RHEL on the master member of the cluster and our web
server, but it would be unpractical to pay a license for each slave, or
for each personal dekstop / laptop in the group.
Thanks
Alfredo
On Tue, 17 Oct 2006, Lonni J Friedman wrote:
> At the risk of pointing out an unpopular concept, if you're this
> sensitive to unstable kernels leaking into your production
> environment, why are you running FC at all? RHEL exists explicitly
> for people with your needs.
>
> On 10/17/06, Alfredo Ferrari <list at pceet030.cern.ch> wrote:
>> You are right, but when you have to deal with 50+ machines you tend to
>> automatize updates for obvious reasons.
>>
>> Further, no warning has been issued, I just saw the thread by chance.
>> Even on my personal machines where I do things by hand, I downloaded the
>> new kernels yesterday, rebuilt them with some extra patches (ntfs,
>> initrd dsdt for my buggy laptop etc), installed them together with the old
>> one, tested for one day (they are under heavy load anyway), tested
>> also (and built the drivers for) hardware acceleration, wireless, dumb
>> Conexant modem etc etc. and eventually I deinstalled the old kernel just
>> before reading that message.
>> It is not a big deal to reinstall the previous kernel on those machines,
>> crossing fingers that nothing happened to the 1k partitions,
>> much tougher on the cluster, but I really miss the rationale of having a
>> kernel update in those conditions (the bug was know as far as I
>> understand).
>> Alfredo
>>
>> On Tue, 17 Oct 2006, Jose Celestino wrote:
>>
>> > Words by Alfredo Ferrari [Wed, Oct 18, 2006 at 12:05:14AM +0200]:
>> >> Seriously, I believe this is a big issue. Let me summarize:
>> >>
>> >> a) there was a kernel update for FC5
>> >> b) this kernel has a known bug which could results in corrupting
>> >> ext3 filesystems with 1k block size under heavy load
>> >> c) ... nevertheless it has been pushed out with no special warning
>> >> d) pratically all /boot partitions are ext3 1k (anaconda generated)
>> >> e) many partitions on old machine upgraded from previous versions are
>> >> ext3 1k as well
>> >> f) experienced users could have much bigger partitions manually
>> >> generated with 1k block size for their own fun/reasons/optimization
>> (I
>> >> personally have counted already 400 GBytes of 1k, ext3, partitions
>> >> just on my personal laptop, desktop and associated backup disks
>> >> excluding /boot ones). In my case, most of the 1k partitions are such
>> >> because they are subject to heavy loads with many small files
>> >>
>> >> What was the rationale for releasing an official kernel update under
>> such
>> >> dangerous conditions? Just "anaconda doesn't generate 1k partitions (not
>> true BTW)"? I still believe Linux is not (yet) Windows
>> >>
>> >
>> > Thank god no. But the growing user base is more and more windows like.
>> >
>> > Thankfuly many of us know that you shouldn't push every and all upgrade
>> to
>> > production servers as soon as it comes. And also thankfuly there are
>> > those who do just that.
>
>
>
--
+----------------------------------------------------------------------------+
| Alfredo Ferrari || Tel.: +41.22.767.6119 |
| C.E.R.N. || Fax.: +41.22.767.7555 |
| European Laboratory for Particle Physics|| |
| AB Division / ATB Group || e-mail: |
| 1211 Geneva 23 || Alfredo.Ferrari at cern.ch |
| Switzerland || Alfredo.Ferrari at mi.infn.it |
+----------------------------------------------------------------------------+
More information about the fedora-list
mailing list