I didn't say that I am surprised.
In fact I am surprised because I observed the fact that the slope is
the same as we have seen in previous upgrades.
I am surprised because this time we set a goal and tried to reach it in
time, but this seems to have had not the slightest effect on the
upgrade rate.
Maybe we discovered a law of nature here.
No matter how hard you push, or how small the changes, you can't exceed
an average upgrade rate of 3 sites/day
CERN's delayed entry into the circle of sites that have made it up to
LCG-2_4_0 has, as Laurence pointed out, many reasons and we have to
improve,
our process to do an upgrade of the production services is far too
slow, especially when compared with the ability of the fabric
management tools used
for upgrading the WN farm. Next time we'll be better.
markus
On Apr 27, 2005, at 4:03 PM, Fotis Georgatos wrote:
> Hi Markus,
>
> Markus Schulz <[log in to unmask]> wrote:
>
>> status of LCG2 release 2_4_0 ( now with a color scheme that allows
>> Windows users to see the scale!)
> [...]
>>> According to our information system we have 136 sites and 50 are on
>>> the latest version.
>>> The VERY-VERY-VERY old contains all sites that are neither 2.4.0,
>>> 2.3.1 or 2.3.0.
>>> Note, we have still more than 20 of them.
>
> It is interesting to note that the curves of "upgrade rates"
> show some striking similarity, with comparable lead-in slope.
> I think this translates roughly to 3 sites per day, at best,
> so the current number of 60 fresh sites should be no surprise at all.
>
> There is certainly something to learn from that, and this comment
> is particularly prompt now that the 3 week time window for 2.4.0
> expires.
>
> I noticed that the CERN site (LXBATCH) does update right at the end,
> too.
> Now, should this be considered a good or a bad practice? And by whom?
> ;-)
>
> --
> echo "sysadmin know better bash than english" | sed s/min/mins/ \
> | sed 's/better bash/bash better/' # Yelling in a CERN forum
>
>
************************************************************************
*******
Markus Schulz
CERN IT
|