Jimmy (Jimbo) Wales a écrit :
Ashar Voultoiz wrote:
Yes. Not enough cpu / disk space to generate
statistics. iirc the last
attempt resulted in a total block of the whole cluster. Hopefully one
server will be dedicated to handling logs in a near future and webalizer
stats will then be build again.
I strongly support the idea of a machine for handling logs,
statistics, and research into those things. I think we could learn a
lot about how the community really functions by studying logs, and the
traffic stats are important for us to understand and project our
growth patterns.
My question to wikitech-l: what sort of machine should we use for
this? I am thinking that a fully loaded dual Opteron is overkill, but
a typical apache is too small?
Probably we'd want to have a lot of disk space, possibly RAID 5 as the
best balance between storage space, redundancy, and speed. (We don't
want to lose a huge chunk of data to a bad hard drive, but on the
other hand we don't need absolute speed either.)
Probably we'd want a decent CPU, but it doesn't have to be top-notch,
since these are batch jobs and the machine should not be doing
anything else anyway.
Your thoughts?
--Jimbo
_______________________________________________
Wikitech-l mailing list
Wikitech-l(a)wikimedia.org
http://mail.wikipedia.org/mailman/listinfo/wikitech-l
A storage server like
http://www.siliconmechanics.com/i1547/serial-ata-storage-server.php
would be nice I think. We could use it to store logs, backup, compute
logs, compute stats
Not really need to be fast a config example ::
Dual Xeon 2,4Ghz, 1GB
12 seagate 400GB drive (not sure we need all this space)
raid 5 with 1 (2?) hot spare, that mean with 1 hot spare 10x400GB = 4To
of space to store logs and backup (backup take more and more place every
day, commons yet eat ~6GB)
cost : 9359$
with 12x250GB, cost fall to 6335$ , with 1 hot spare 10x250GB = 2,5To
My 2 cents