There is few reason for high CPU load in dom0 for XCP. Note: any reason
except diskIO is bad sign.
First: high IO utilization from domUs via (f.e. ISCSI). I found that
really high load from guests can really load domU (above 1.5Gb/s). This
is unavoidable (may be hardware accelerated ISCSI HBA? Not test them
yet).
Second: Bad sync. If some host lost sync with others, high load can
occur on this host.
Third: Heavy use of xe interface. I''ve tryed to make some kind of
simple
cloud control system based on ssh+xe syntax. xe vm-list (or any other
*-list) couse extreme load on slave host (where command issued) and
master host (which one really serve request). Main symptom: xapi eat up
to 30-70% of CPU. Result: do not use xe interface for any automated and
often running tasks. Use xenApi, use event subscribe for state
replication. Try avoid any requests which one requied lot of operation
for XenAPI server.
Forth: Migration cause stress on stunnel due high data transfer, so, as
rule of thumb: one migration (accept or send) per host in one time.
As I found it tests, normal load for dom0 is about 10-30% for half-idle
guest activity.
В Вск, 22/08/2010 в 15:28 -0400, Lei Lu пишет:> Hi all,
>
> I happened to run into a infinite loop on my guest domain, I found by
> xentop or xenmon that my guest domain cpu is 100%. However,
> surprisingly, domain-0 cpu utilization also around 100%. I am very
> confused and do not know why it is this case, could anybody tell me
> the reason or something I should look into? Thank you in advance.
>
> Best,
> Lei
>
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users
_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users