Subject: Re: xenbsd /netbsd: xennetback: can't get new mcl pages (0) (Maybe
To: <>
From: Anzi <anzi@dnainternet.net>
List: port-xen
Date: 03/12/2007 09:18:28
Manuel Bouyer wrote:
> On Tue, Mar 06, 2007 at 01:16:19PM +0200, Anzi wrote:
>   
>> Manuel Bouyer wrote:
>>     
>>> On Tue, Mar 06, 2007 at 08:18:53AM +0200, Anzi wrote:
>>>  
>>>       
>>>>> How much free ram do you have ? xm info should tell you.
>>>>> It's possible that There are too much ram allocated to domUs, would cause
>>>>> problem when dom0 needs some pages to perform I/O with the domUs.
>>>>>
>>>>>
>>>>>      
>>>>>           
>>>> 433 Mb of free memory, I think that is enough? Xen version is from 
>>>> pkgsrc, 3.0.3 I think.
>>>>    
>>>>         
>>> 433 should be enough. Do you have some linux domain that can baloon and
>>> eat all this free ram ?
>>>
>>>  
>>>       
>> Hmm, both domains are fedora core 6 linuxes (originally it was supposed 
>> that other is  guest is netbsd, netbsd doesn't all required finnish LC_* 
>> localisation implemented). How they can eat all free ram? What exactly 
>> does the max memory spesified in guest conf mean?
>>     
>
> AFAIK it's really a hard limit for a domU. dom0 can grow higher;
> xm top will tell you the real usage. It would be interesting to run
> xm top while you get the xennetback error.
>
>   
I think I found out what caused the problem. I had setup cron job to 
make a daily and weekly backups of root filesystem. Weekly job copies 
dump image from /backup/dumps/ (different disk) to NFS share on tape 
server and daily job produces dumps to /backup/dumps. Cron timing was 
setup @weekly and daily was 10 minutes after midnight. So the daily job 
was overwriting the image as it was copied to NFS volume. I don't know 
why it failed with the error but that was the cause of problem.
This weekend it worked ok.

-anzi-