On 09/18/2012 12:16 AM, Prem Karat wrote:
On 09/13/12 03:41pm, Chris Evich wrote:
On 09/13/2012 01:16 PM, Prem Karat wrote:
On 09/12/12 02:34pm, Chris Evich wrote:
Thanks for the fix! Your reasoning sounds good to me. I have just
two questions...
On 09/12/2012 01:36 PM, Prem Karat wrote:
virsh freecell has libvirt=on& libvirt=off as the main variants and hence
evey test under freecell will include either on of them as the parameter.
Hence while preparing libvirt service for every test, this check can be
avoided.
# Prepare libvirtd service
check_libvirtd = params.has_key("libvirtd")
if check_libvirtd:
Params will always have either libvirtd as on or off as variants show
below in subtests.cfg.sample
variants:
- libvirton:
libvirtd = "on"
- libvirtoff:
libvirtd = "off"
status_error = "yes"
Cleaning up to avoid the unnecessary check and introducing a check to
see the status of libvirt daemon and if its not running, will start the
service on "libvirtd = on" variant.
As much as I try to catch them in patches, it's conceivable that a
prior virsh_* test had a problem and failed to recover libvirtd
service. In this case, if virsh_freecell happens to be the
following test, it may hide a bug or test problem. Assuming this is
unlikely event, do you think it's worth worrying about?
What kind of problem are you referring here? Assuming Test Failures, that should
be captured by the individual tests.
I wouldn't worry about that because all we are trying to do is to ensure that
libvirtd service is running for any virsh_commands to execute.
Sorry for not being clear, "The Problem" is, if we don't have your
fix on every virsh test, when libvirtd is ever in wrong state, all
subsequent virsh_* tests will fail.
Second, if this patch is implemented. What do you think if I were
to make it the "standard" by putting this libvirtd
check/start/restart/stop stuff into a common place, like virt_env
pre/post process?
i.e. it seems nearly all the virsh_* tests are taking these same
steps, it would be nice to have uniform behavior.
Yes I agree with that. Preparing libvirt during pre process and recovering post
process would be the ideal thing to have.
Okay good, let's take this approach so we don't have a large
collection of virsh_* tests all doing different things with libvirtd
service :S Would you like to attempt putting this on pre-process?
No worries if not, I'll take a look at it next week once we get the
big virt code split stuff done (Monday I believe is the day).
Yes, am working on to get this in pre-process& post-process step. I'll get back
to you on this.
Prem,
On a somewhat related note, take a look at
https://github.com/autotest/autotest/issues/541 This is a more generic
way of handling the situation, but with extra powers - tests can
change/override the pre/post behavior as needed. Please feel free to
take on this issue if it's something that perks your interest.
--
Chris Evich, RHCA, RHCE, RHCDS, RHCSS
Quality Assurance Engineer
e-mail: cevich + `@' + redhat.com o: 1-888-RED-HAT1 x44214
_______________________________________________
Autotest-kernel mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/autotest-kernel
