sagenb

El viernes, 13 de enero de 2017, 19:34:15 (UTC+1), Dima Pasechnik escribió:
>
>
>
> On Friday, January 13, 2017 at 5:30:04 PM UTC, Enrique Artal wrote:
>>
>> Putting limits in /etc/security/limits.conf (or in files in limits.d) 
>> works right up to Sage 7.3. Namely, if a user performs a strong computation 
>> (memory or CPU time), the system stops the computation when the limit is 
>> reached; usually one needs to quit the worksheet, but it is possible to 
>> reuse the notebook. With 7.4 and 7.5, when the limit is reached the 
>> notebook becomes unusable and the only possibility to work is to kill and 
>> restart it. Some change between 7.3 and 7.4 may cause it.
>>
>
> "the notebook"? Which one? sagenb, or jupyter?
>  
>
>>
>>
>> El domingo, 27 de noviembre de 2016, 21:55:06 (UTC+1), Enrique Artal 
>> escribió:
>>>
>>> It seems to work now with the ulimits for the server_pool users. If they 
>>> become too strict, we (maybe more precisely MIguel Marco) will try the 
>>> worker user approach. We will let know. Thanks for the help!
>>>
>>> El domingo, 27 de noviembre de 2016, 21:23:33 (UTC+1), Nils Bruin 
>>> escribió:
>>>>
>>>> On Sunday, November 27, 2016 at 3:04:48 AM UTC-8, Enrique Artal wrote:
>>>>>
>>>>> Thanks, As you say, it would be better something more direct, but your 
>>>>> approach is a strong improvement for my needs. 
>>>>> By the way, I changed in our experimental notebook 7.4 -> 7.3 and the 
>>>>> limits work: they stop the process and the notebook is still running.
>>>>>
>>>>
>>>> for sage 7.5beta(?) setting ulimits does have effect: with
>>>>
>>>> sh$ ulimit  -v 10000000
>>>> sh$ sage -c 'L=[1]
>>>> for i in [1..1000]:
>>>>   L = L+L
>>>>   print i'
>>>>
>>>> I get a memory error after "28" has been printed (and without it, it 
>>>> continues longer), and if I take the bound much lower sage will not even 
>>>> start.
>>>>
>>>> So if you configure the "worker" user to have such a ulimit, I'd expect 
>>>> memory problems to be significantly reduced. People who try to use more 
>>>> memory should see their kernel die before it's causing problems for other 
>>>> people.
>>>>
>>>> Given that there's no way of controling which notebook user gets mapped 
>>>> to which worker uid, I don't think there's much mileage to be had from 
>>>> configuring multiple worker uids (other than having them on multiple 
>>>> machines to load-balance a little bit).
>>>>
>>>>

-- 
You received this message because you are subscribed to the Google Groups 
"sage-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at https://groups.google.com/group/sage-support.
For more options, visit https://groups.google.com/d/optout.

Reply via email to