Re: [libvirt] [test-API][PATCH] balloon_memory: add time break before dump xml

On 11/27/2012 04:36 PM, Guannan Ren wrote: > On 11/27/2012 03:02 PM, Wayne Sun wrote: >> the xml dumped after setMemory is not accurate due to time >> delay, so take 10 sec sleep before dump to show the right >> xml info > > Why it is necessary to sleep for a while, I don't think > adding a sleep > is a good way to solve problem. > Please try to figure out the root cause. > > Guannan > DB recently change the code of get domain memory balloon value as in patch: http://www.redhat.com/archives/libvir-list/2012-May/msg00871.html If QEMU supports the BALLOON_EVENT QMP event, then we can avoid invoking 'query-balloon' when returning XML or the domain info. And in qemu patch: http://lists.gnu.org/archive/html/qemu-devel/2012-05/msg02833.html DB also emphasis in description: It is important to note that events are only discarded when they are obsoleted by a newer event. So an application is guarenteed to see the final balloon event, with at worst a 1 second delay. So when after do memory change, for dump domain xml to get balloon value, it will be controlled in 1 second delay. When i check with QMP with query-events commands on qemu-kvm-rhev-0.12.1.2-2.337.el6.x86_64: { "execute": "query-events"} {"error": {"class": "CommandNotFound", "desc": "The command query-events has not been found", "data": {"name": "query-events"}}} and this also in my log, so my qemu did not support this BALLOON_EVENT and using query-balloon to get current memory, and indeed 'query-balloon' is spotted in my log. This back to # vim src/qemu/qemu_driver.c +5330 /* Don't delay if someone's using the monitor, just use * existing most recent data instead */ so most recent data will be returned with query-balloon, that means if we query too fast, the result will not be accurate. I think add little time sleep is reasonable here for now, but after qemu support BALLOON_EVENT, then the sleep will be no use. So, what do you think?