Xen users - Jan 2011 - XM sched-credit

Hi,

I have modified the sched-credit for one virtual machine, using xm 
sched-credit -d 0 -w 60000, and the next day Nagios reported some 
problems on Dom0. Two plugins, that are checking the firewall and 
smartctl reported "socket timeout after 10 seconds". When I issued the
two scripts, one of them took 5 seconds to run, and the other one a few 
minutes. Here is the first command that took 5 seconds:

/sbin/iptables-save|grep -v ''^\*'' |grep -v
''^#'' |grep -v ''^COMMIT''|grep
-v ''^:''|wc -l

When straced, it pauses for a long time in wait4(). When issued several 
times, it waits a long time on the third or forth grep. The second one 
does the same in read(). The load on the dom0 is 0.02, ulimit is ok,and 
there is no waiting time for the cpu. Also, the virtual machines are 
running smoothly, without high IO, memory or CPU.

Could the problem be generated by the scheduler, that allocated a small 
weight for the dom0?

Thx in advance,

Paul PISCUC

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

Yes, this might be a solution. The problem is that on the Dom0 I have no 
application running, except for the firewall, and some nagios plugins, 
that are interrogated by the nagios server. When one guest requires a 
lot of CPU, the nagios server sends an alert that the Dom0 is not 
responding, although is has no load or other issues. This happens almost 
once a week, and the guest OS is used for a database server.

  What would be a better solution ... to pin a cpu for dom0, to change 
the weight for dom0, or to continue using the same sched-credit, and 
tune the nagios server to wait for Dom0 longer?

Thx in advance.
> yes. seems you allocated all cpu ressources away from dom0.
> can you try pinning the vm to one cpu and dom0 to another?
>
> Florian
>
> 2011/1/14 Paul Piscuc<paul.piscuc@appnor.com>:
>> Hi,
>>
>> I have modified the sched-credit for one virtual machine, using xm
>> sched-credit -d 0 -w 60000, and the next day Nagios reported some
problems
>> on Dom0. Two plugins, that are checking the firewall and smartctl
reported
>> "socket timeout after 10 seconds". When I issued the two
scripts, one of
>> them took 5 seconds to run, and the other one a few minutes. Here is
the
>> first command that took 5 seconds:
>>
>> /sbin/iptables-save|grep -v ''^\*'' |grep -v
''^#'' |grep -v ''^COMMIT''|grep -v
>> ''^:''|wc -l
>>
>> When straced, it pauses for a long time in wait4(). When issued several
>> times, it waits a long time on the third or forth grep. The second one
does
>> the same in read(). The load on the dom0 is 0.02, ulimit is ok,and
there is
>> no waiting time for the cpu. Also, the virtual machines are running
>> smoothly, without high IO, memory or CPU.
>>
>> Could the problem be generated by the scheduler, that allocated a small
>> weight for the dom0?
>>
>> Thx in advance,
>>
>> Paul PISCUC
>>
>> _______________________________________________
>> Xen-users mailing list
>> Xen-users@lists.xensource.com
>> http://lists.xensource.com/xen-users
>>
>
>

-- 
Best regards,
Paul PISCUC
System Administrator - Appnor MSP S.A.
US: (+1) 650 336 57 50
RO: (+4) 021 569 46 56
http://www.appnor.com


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users