Jaroslaw Polok wrote:
> Randy Merritt wrote:
> 
>>  /etc/cron.daily/yum.cron:
>>
>> Yum crontab starting
>>   Sleeping for 153 minutes
>>   No other yum process running.
>> ********** Starting ***** errata ***** STARTING ********
>>    Creating Config File
>> *********START** errata YUM.CONF FILE **START**********
>> .  .  .
>>
>> Can someone illuminate why yum.cron (or cron itself?) is sleeping for 
>> 153 minutes?
> 
> 
> In order not to kill servers serving you updates yum cron job
> sleeps for a random time (not longer than 180 minutes - see
> /etc/cron.daily/yum.cron):
> 
> If every  SL user would update at the same time ...
> Connie would have to have a real http serving farm at 
> scientificlinux.org ;-)
> 
> Cheers
> 
> Jarek

Exactly.
I don't know if you've seen the stat's, but there are currently 12,000 
machines doing their updates from us.  Currently, by using this delay, 
and the fact that they are distributed around the world, the load on the 
machine is fairly consistant.

(side note: we are working on a program to identify 'clusters' of 
machines.  If you have several machines getting their updates from us, 
we encourage people to mirror us and point their yum at their mirror. 
This saves our bandwidth, as well as gives you faster responses.)

If everyone took the randomizing out of their cron job's, we'd see huge 
spikes of traffic each hour, and then this e-mail wouldn't be about long 
sleeps, it would be about long lags when doing updates.

At Fermilab we have our own servers we point our yum at (so we arn't 
hitting the main server) and it is usually quite obvious when even a 
cluster of 100 machines takes that randomizing out.

Troy
-- 
__________________________________________________
Troy Dawson  [log in to unmask]  (630)840-6468
Fermilab  ComputingDivision/CSS  CSI Group
__________________________________________________