[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] crash in csched_load_balance after xl vcpu-pin



On Wed, 2018-04-11 at 08:23 +0200, Olaf Hering wrote:
> It turned out that I had a typo all the time in my template, it used
> 'cpu=' rather than 'cpus='. On this system none of this works:
> #pus="node:${node}"
> cpus="nodes:${node}"
> #pus="nodes:${node},^node:0"
> #pus_soft="nodes:${node},^node:0"
> 
> Only 'cpus=node:1' or 'cpus=nodes:1' works, cpus=node:2 or node:3
> does
> not.
> There is room for domUs:
> numa_info              :
> node:    memsize    memfree    distances
>    0:     30720       1912      10,21,21,21
>    1:     28672      22355      21,10,21,21
>    2:     24576      24502      21,21,10,21
>    3:     32768      31760      21,21,21,10
> 
So, now, when you say 'does not work', do you mean 'domain creation is
aborted with errors' or 'domain is created, but memory is not where it
should be'.

IAC, here, when using `xl vcpu-pin':

root@Zhaman:/home/dario# xl vcpu-pin 1 all node:0-1,^nodes:0,4-7,^5
root@Zhaman:/home/dario# xl vcpu-list 1
Name                                ID  VCPU   CPU State   Time(s) Affinity 
(Hard / Soft)
leap15                               1     0   14   -b-      15.5  4,6-15 / all
leap15                               1     1   11   -b-       4.7  4,6-15 / all
leap15                               1     2   12   -b-       4.6  4,6-15 / all
leap15                               1     3   14   -b-       4.3  4,6-15 / all
leap15                               1     4   14   -b-       5.8  4,6-15 / all
leap15                               1     5    8   -b-       4.5  4,6-15 / all
leap15                               1     6   10   -b-       4.3  4,6-15 / all
leap15                               1     7    9   -b-       3.7  4,6-15 / all

If I shut the domain down, and re-create it with cpus="..." and
cpus_soft="...":

root@Zhaman:/home/dario# cat vms/hvm/leap15.cfg |grep -e "cpus[=|_]"
cpus="node:0,^4-6,nodes:1,^12,^14"
cpus_soft="nodes:0,^node:1"
root@Zhaman:/home/dario# xl vcpu-list 2
Name                                ID  VCPU   CPU State   Time(s) Affinity 
(Hard / Soft)
leap15                               2     0   13   -b-      17.3  
0-3,7-11,13,15 / 0-7
leap15                               2     1    7   -b-       7.4  
0-3,7-11,13,15 / 0-7
leap15                               2     2    0   -b-       6.0  
0-3,7-11,13,15 / 0-7
leap15                               2     3    2   -b-       7.9  
0-3,7-11,13,15 / 0-7

And from `xl debug-key u':
(XEN) [ 3841.835310] Domain 2 (total: 1044554):
(XEN) [ 3841.844555]     Node 0: 1044554
(XEN) [ 3841.844559]     Node 1: 0

which is fine, because soft-affinity is used, if it is explicitly specified.

So, I'd say that all seems to work fine, even using nodes and then only
a single digit, or using node and then a range. And also doing all the
various set manipulations, like "nodes:0,^node:1".

I really am not sure what the issue could be, there at your side...

Dario
-- 
<<This happens because I choose it to happen!>> (Raistlin Majere)
-----------------------------------------------------------------
Dario Faggioli, Ph.D, http://about.me/dario.faggioli
Software Engineer @ SUSE https://www.suse.com/

Attachment: signature.asc
Description: This is a digitally signed message part

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.