[Ovirt-devel] [PATCH] Test System Updates

[Mohammed Morsi]

Perry N. Myers wrote:
> Mohammed Morsi wrote:
>> Took these fixes, and our discussion about how the 2nd stage 
>> autobuild is unnecessary and came up with a simpler autobuild script 
>> / kickstart that runs the 2nd stage tests (eg site accessibility and 
>> rake tests) through ssh. When testing this, I noticed some of the 
>> rake tests failed, I believe it due to recent code changes, and will 
>> look at resolving them once this autobuild stuff is committed. In 
>> addition to the rrd test data generation and usage code in the 
>> patched, I began looking into how we can add some test data to the 
>> rh-dirsrv ldap server running on the ovirt wui appliance. I'm not a 
>> 100% sure yet, but I might be able to add a ldif file with test data 
>> to the appliance via a kickstart script and then configure the server 
>> / ovirt to use a path that points to it.
>>
>>   -Mo
>>
>>
>
> Comments inline...  I think Alan has made some of these comments 
> already, but just to make sure...
>
>> -#run tests
>> -cd wui/src/
>> -rake db:migrate
>> -rake test
>> +echo "Running oVirt Autobuild"
>> +
>> +# create appliance
>> +./build-all.sh -ac
>
> With Alan's changes to build-all, pungi no longer init's by default.  
> I think we should init manually once on harpoon, and then do:
>
> ./build-all.sh -acp update
>
> to update pungi for each run.
>
>> +
>> +while : ; do
>> +    status=`virsh domstate ovirt-appliance`
>> +    if [ -f /etc/libvirt/qemu/ovirt-appliance.xml -a "$status" = 
>> "shut off" ]; then
>> +        break
>> +    fi
>> +    sleep 10
>> +done
>
> No need for a loop here since with Alan's patch appliance creation is 
> synchronous.
 From what I can tell (unless this has changed w/ the latest patch, 
which I thought it didn't) the creation of the appliance on the host 
side is synchronous, but when the appliance starts up (eg build all says 
'run virt-viewer to view output') and Fedora being installed, the 
build-all command returns. I need to wait until this install process is 
a 100% complete, before continuing and thus the loop is neccessary (once 
again, basing this on what I observed, correct me if I am wrong).
>
>> +
>> +# start appliance
>> +virsh start ovirt-appliance
>> +
>> +# wait until started
>> +sleep 180
>
> Hmm.  Sleep 180 is fine for an initial wait (since we know it'll be a 
> few minutes for the appliance to come up) but below you just 
> immediately start ssh'ing into the appliance.  Instead, do a looping 
> ssh test after the sleep.  Once the ssh test succeeds, then continue 
> on with the rest of the script.
The ssh test is a good idea, sounds like it might even mitigate the need 
for the sleep.
>
>> +
>> +# the appliance's ssh key changed
>> +rm -f /root/.ssh/known_hosts
>
> In your ssh commands below, just use:
> SSH_OPTS="-o StrictHostKeyChecking=no -o UserKnownHostsFile=/dev/null"
> This way you don't need to muck with root known_hosts file.
>
>> +
>> +echo "Running the wui tests"
>> +ssh root at 192.168.50.2 "curl -i --negotiate -u : 
>> management.priv.ovirt.org/ovirt/ | grep \"HTTP/1.1 200 OK\" && \
>> +           cd /usr/share/ovirt-wui/ && rake test"
>> diff --git a/wui-appliance/common-pkgs.ks b/wui-appliance/common-pkgs.ks
>> index 34546a7..d694332 100644
>> --- a/wui-appliance/common-pkgs.ks
>> +++ b/wui-appliance/common-pkgs.ks
>> @@ -21,4 +21,7 @@ xorg-x11-xauth
>>  virt-viewer
>>  cobbler
>>  bind-utils
>> -
>> +perl-Test-AutoBuild
>> +perl-Test-AutoBuild-git
>> +perl-Class-MethodMaker
>> +curl
>
> If 2nd stage autobuild is unnecessary on the appliance, why are these 
> packages still in the appliance kickstart?
I meant to just include curl to test that the appliance wui was running, 
I must've forgot to remove the AutoBuild stuff.
>
> Perry
>
  Mo