On Wed, Feb 27, 2013 at 03:26:24PM +0100, Michal Privoznik wrote:
On 27.02.2013 12:20, Daniel Veillard wrote:
> I have just tagged git and pushed the tarball. The rpms for F17
> are at the usual place too:
>
ftp://libvirt.org/libvirt/
>
> I gave a try to the set of rpms and this seems to work fine for
> relatively simple tests.
> Please give it more testing and let's keep change to git to
> bug fixes to minimize risks for 1.0.3.
> If everything goes well I will push 1.0.3 on Friday,
>
> thanks !
>
> Daniel
>
Maybe we want to postpone the release a bit; I was trying to find out how drop of qemu
driver lock speeds up parallel startup of multiple domains, however - it is not working.
I've created small test program - test4.c and run it against current HEAD. It takes
arguments - domain names - which it spawns a separate thread per each domain and
repeatedly starts and destroys domains. See code for more.
Then, I've created 20 copies of diskless domain (attached as well): test1..test20 and
run my test4 program over them:
./test4 $(for ((i=1; i<21; i++)); do echo -n "test$i "; done)
I had to tweak max_client_requests=20 to really run those 20 requests in parallel.
However, what I am getting instead of nice measuring of speedup is:
[9722] Starting test1 (round 20) ...[9725] Starting test4 (round 20) ...[9724] Starting
test3 (round 20) ...[9723] Starting test2 (round 20) ...[9726] Starting test5 (round 20)
...[9727] Starting test6 (round 20) ...[9728] Starting test7 (round 20) ...[9729] Starting
test8 (round 20) ...[9730] Starting test9 (round 20) ...[9731] Starting test10 (round 20)
...[9732] Starting test11 (round 20) ...[9733] Starting test12 (round 20) ...[9734]
Starting test13 (round 20) ...[9735] Starting test14 (round 20) ...[9736] Starting test15
(round 20) ...[9737] Starting test16 (round 20) ...[9738] Starting test17 (round 20)
...[9741] Starting test20 (round 20) ...[9739] Starting test18 (round 20) ...[9740]
Starting test19 (round 20) ...libvir: error : An error occurred, but the cause is
unknown
Unable to start domain test12libvir: error : An error occurred, but the cause is
unknown
Unable to start domain test14libvir: error : An error occurred, but the cause is
unknown
Unable to start domain test7libvir: error : An error occurred, but the cause is unknown
Unable to start domain test1libvir: QEMU Driver error : Requested operation is not valid:
domain is not running
Unable to destroy domain test12[9733] Destroying test12 ...[9728] Destroying test7
...[9735] Destroying test14 ...[9733] Starting test12 (round 19) ...libvir: QEMU Driver
error : Requested operation is not valid: domain is not running
Unable to destroy domain test7[9728] Starting test7 (round 19) ...libvir: QEMU Driver
error : Requested operation is not valid: domain is not running
Unable to destroy domain test14[9735] Starting test14 (round 19) ...libvir: error : An
error occurred, but the cause is unknown
Unable to start domain test18libvir: QEMU Driver error : Requested operation is not
valid: domain is not running
Unable to destroy domain test1[9722] Destroying test1 ...[9722] Starting test1 (round 19)
...libvir: error : An error occurred, but the cause is unknown
Unable to start domain test8libvir: error : An error occurred, but the cause is unknown
Unable to start domain test15[9725] Done
etc.
Unfortunately, I caught this after freeze.
No problem, freeze is there to find bugs, you found a serious bug,
let's fix it before the release.
[...]
//sleep(20);
sleep(1);
does sleep(20); there avoid the problem ? There is a failure in guest
start but we don't get a proper description
There is no urging deadline, except well for the pending patches
waiting for the end of the freeze, if we can fix this by Monday, then
that would be IMHO just fine. If on Monday we still don't have a
solution let's discuss the status of the analysis,
okay ?
Daniel
--
Daniel Veillard | Open Source and Standards, Red Hat
veillard(a)redhat.com | libxml Gnome XML XSLT toolkit
http://xmlsoft.org/
http://veillard.com/ | virtualization library
http://libvirt.org/