Warly wrote:

John Allen <[EMAIL PROTECTED]> writes:



On Thursday 06 November 2003 20:35, Blindauer Emmanuel wrote:


Le Jeudi 06 Novembre 2003 18:14, Robert Fox a �crit :


Maybe someone should be a mirror monitor and be the main point of
contact with the mirror owners when something goes wrong!


http://extasia.u-strasbg.fr/~blindaue/mirrors.php



You don't do uninett.no

Also it would be better to test the mirrors against the hdlist/synthesis of the master site (wherever that is)



updates every 2 hours the status of mirrors, looking into hdlist.cz and
looking for the files inside to be present in the directory of the ftp.

But I can't test more than that.



I tried to have a more efficient rsync process to our mirror reference machine (I decrease contrib frequency mirroring to test if main only can be kept correctly up to date).

Maybe we will have to switch to a every 2 hour frequency if one hour
delay is too short for mirror to catch on.

I'm sorry, but this doesn't sound right.

I'm behind a crappy cable connection and I'm capable of pulling a good mirror, every hour, from kenobi. If kenobi was updated every 10mins, I would probably sync every 10 mins.

I can't imagine why cooker.mandrakesoft.com shouldn't be able to do the same...

IMHO cutting down on the frequency won't help, it'll just worsen the problem. What you need to do is make sure that you don't mirror the same thing twice --> i.e. don't start a second mirror process if the first one is still running.

I'm running the following cronjobs for my mirrors:
0 0-22/2 * * * if [ `ps -u $UID | grep all.sh | wc -l` = 0 ]; then ~/bin/all.sh cooker upload >/dev/null 2>&1 ; fi
0 1-23/2 * * * if [ `ps -u $UID | grep all.sh | wc -l` = 0 ]; then ~/bin/all.sh cooker >/dev/null 2>&1 ; fi

on the even hours packages are uploaded (ftpcooker / ftpcontrib is invoked) on the odd hours they aren't.

If the "all.sh" script is found running under the current user, then a new one doesn't start. Increasing the frequency will make the mirror more up2date and decrease the load on it --> if you need to tranfer 100Mb, it's better to do it 12* 8.125Mb spread over the hour than 1* 100Mb at :00. Again, just make sure you don't two "all.sh" running.

please take this into consideration.

regards,

Stefan

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature



Reply via email to