On Wed, Feb 6, 2019 at 11:05 AM Barak Korren <bkorren@redhat.com> wrote:


בתאריך יום ד׳, 6 בפבר׳ 2019, 12:25, מאת Simone Tiraboschi ‏<stirabos@redhat.com>:


On Wed, Feb 6, 2019 at 11:17 AM Barak Korren <bkorren@redhat.com> wrote:


On Wed, 6 Feb 2019 at 11:57, Simone Tiraboschi <stirabos@redhat.com> wrote:


On Wed, Feb 6, 2019 at 10:44 AM Barak Korren <bkorren@redhat.com> wrote:


On Wed, 6 Feb 2019 at 11:34, Simone Tiraboschi <stirabos@redhat.com> wrote:


On Wed, Feb 6, 2019 at 10:23 AM Barak Korren <bkorren@redhat.com> wrote:


On Wed, 6 Feb 2019 at 11:15, Simone Tiraboschi <stirabos@redhat.com> wrote:


On Wed, Feb 6, 2019 at 10:00 AM Dan Kenigsberg <danken@redhat.com> wrote:
On Wed, Feb 6, 2019 at 10:54 AM Simone Tiraboschi <stirabos@redhat.com> wrote:
>
>
>
> On Wed, Feb 6, 2019 at 9:45 AM Dan Kenigsberg <danken@redhat.com> wrote:
>>
>> On Wed, Feb 6, 2019 at 10:16 AM Simone Tiraboschi <stirabos@redhat.com> wrote:
>> >
>> >
>> >
>> > On Tue, Feb 5, 2019 at 7:07 PM Dafna Ron <dron@redhat.com> wrote:
>> >>
>> >> Hi,
>> >>
>> >> Please note that ovirt-ansible-hosted-engine-setup has a versioning problem with the package and is causing bootstrap to fail for upgrade suite [1]
>> >>
>> >> This is effecting all projects, its been reported to the developers and should be fixed as soon as possible.
>> >>
>> >> you can view CQ status here:
>> >> https://jenkins.ovirt.org/view/Change%20queue%20jobs/job/ovirt-master_change-queue-tester/
>> >>
>> >> [1] http://pastebin.test.redhat.com/708086
>>
>> It is unfair to refer to an internal pastebin here. It is also not
>> very sensible, as it is quite short.
>>
>> 2019-02-05 11:23:51,390-0500 ERROR
>> otopi.plugins.otopi.packagers.yumpackager yumpackager.error:85 Yum
>> [u'ovirt-hosted-engine-setup-2.3.5-0.0.master.20190205110929.gitfdbc215.el7.noarch
>> requires ovirt-ansible-hosted-engine-setup >= 1.0.10']
>> 2019-02-05 11:23:51,390-0500 DEBUG otopi.context
>> context._executeMethod:142 method exception
>> Traceback (most recent call last):
>>   File "/tmp/ovirt-6fV8LBWX5i/pythonlib/otopi/context.py", line 132,
>> in _executeMethod
>>     method['method']()
>>   File "/tmp/ovirt-6fV8LBWX5i/otopi-plugins/otopi/packagers/yumpackager.py",
>> line 248, in _packages
>>     self.processTransaction()
>>   File "/tmp/ovirt-6fV8LBWX5i/otopi-plugins/otopi/packagers/yumpackager.py",
>> line 262, in processTransaction
>>     if self._miniyum.buildTransaction():
>>   File "/tmp/ovirt-6fV8LBWX5i/pythonlib/otopi/miniyum.py", line 920,
>> in buildTransaction
>>     raise yum.Errors.YumBaseError(msg)
>> YumBaseError: [u'ovirt-hosted-engine-setup-2.3.5-0.0.master.20190205110929.gitfdbc215.el7.noarch
>> requires ovirt-ansible-hosted-engine-setup >= 1.0.10']
>> 2019-02-05 11:23:51,391-0500 ERROR otopi.context
>> context._executeMethod:151 Failed to execute stage 'Package
>> installation': [u'ovirt-hosted-engine-setup-2.3.5-0.0.master.20190205110929.gitfdbc215.el7.noarch
>> requires ovirt-ansible-hosted-engine-setup >= 1.0.10']
>> 2019-02-05 11:23:51,413-0500 DEBUG
>> otopi.plugins.otopi.debug.debug_failure.debug_failure
>> debug_failure._notification:100 tcp connections:
>>
>> >>
>> >
>> > The issue is that on github we already have
>> > VERSION="1.0.10"
>> > as we can see in
>> > https://github.com/oVirt/ovirt-ansible-hosted-engine-setup/blob/master/build.sh#L3
>> >
>> > And this has been bumped before the commit that now is reported as broken.
>> >
>> > CI instead is still building the package as 1.0.9 ignoring the commit that bumped the version.
>> > Honestly I don't know how I can fix it if the version value is already the desired one in the source code.
>>
>> I don't see your ovirt-ansible-hosted-engine-setup-1.0.10, only
>> https://plain.resources.ovirt.org/pub/ovirt-master-snapshot/rpm/el7/noarch/ovirt-ansible-hosted-engine-setup-1.0.9-0.1.master.20190129095419.el7.noarch.rpm
>> Not even under "tested":
>> https://plain.resources.ovirt.org/repos/ovirt/tested/master/rpm/el7/noarch/ovirt-ansible-hosted-engine-setup-1.0.9-0.1.master.20190129095419.el7.noarch.rpm
>>
>> Simone, can you doublecheck that its artifacts have been built and
>> have been accepted by the change queue?
>
>
> It has been built here once as 1.0.10:
> https://jenkins.ovirt.org/job/oVirt_ovirt-ansible-hosted-engine-setup_standard-on-ghpush/91/
>
> then on the next commit, CI started building it again as 1.0.9 although in the source code we have 1.0.10 and so this issue.

I don't understand the issue yet (that's not surprising as I do not
know what is that "ghpush" job). Which CI job has built the wrong
version? can you share its logs? who owns it?

In the git log I see:
commit b5a6c1db135d81d75f3330160e7ef4a84c97fd60 (HEAD -> master, upstream/master, origin/master, origin/HEAD, nolog)
Author: Simone Tiraboschi <stirabos@redhat.com>
Date:   Tue Feb 5 10:56:58 2019 +0100

    Avoid using no_log when we have to pass back values to otopi

commit 96974fad1ee6aee33f8183e49240f8a2a7a617d4
Author: Simone Tiraboschi <stirabos@redhat.com>
Date:   Thu Jan 31 16:39:58 2019 +0100

    use dynamic inclusion to avoid tag inheritance

commit 4a9a23fb8e88acba5af4febed43d9e4b02e7a2c5
Author: Simone Tiraboschi <stirabos@redhat.com>
Date:   Thu Jan 31 15:15:04 2019 +0100

    Force facts gathering on partial executions

commit 7428b54a5ba8458379b1a27d116f9504bb830e69
Author: Simone Tiraboschi <stirabos@redhat.com>
Date:   Wed Jan 30 17:01:01 2019 +0100

    Use static imports and tags
    



Version has been bumped to 1.0.10 on commit 7428b54a5ba8458379b1a27d116f9504bb830e69 since it introduces a backward incompatible change and we need to track it.

7428b54a5ba8458379b1a27d116f9504bb830e69 failed CI tests due to an issue on a different package found yesterday.

So 7428b54a5ba8458379b1a27d116f9504bb830e69 got ignored and now CI is building from commit b5a6c1db135d81d75f3330160e7ef4a84c97fd60 (the last one) rebased on something before 7428b54a5ba8458379b1a27d116f9504bb830e69 which is not what we have in git so now, after b5a6c1db135d81d75f3330160e7ef4a84c97fd60 (last commit) the package builds in CI as 1.0.9 although in the code we have 1.0.10 and so the issue.


We never ignore commits, certainly not merged ones...

We can fall back to older builds on system test failures and throw away newer build if we suspect they cause the failure, if which case the builds need to be resubmitted, but this logic happens at the build leve not the commit level, there is no commit reordering or dropping anywhere.


I'd double check the CI code for that:

on git side, in build.sh at commit b5a6c1db135d81d75f3330160e7ef4a84c97fd60 we have
VERSION="1.0.10"
as you can see in
https://github.com/oVirt/ovirt-ansible-hosted-engine-setup/blob/b5a6c1db135d81d75f3330160e7ef4a84c97fd60/build.sh#L3
 
but, as you can see in
https://jenkins.ovirt.org/job/oVirt_ovirt-ansible-hosted-engine-setup_standard-check-pr/138/
CI builds it as 1.0.9 ignoring commit 7428b54a5ba8458379b1a27d116f9504bb830e69
that has been merged before b5a6c1db135d81d75f3330160e7ef4a84c97fd60


I'm not sure if build 138 ran before or after 7428b54a5ba8458379b1a27d116f9504bb830e69 was merged. 
since you merged the PR already, we cannot check what will happen if we rerun the job now.

I suspect you're simply looking at the output of a job that stated running before 7428b54a5ba8458379b1a27d116f9504bb830e69 was merged 


b5a6c1d (HEAD -> master, upstream/master, origin/master, origin/HEAD, nolog) Avoid using no_log when we have to pass back values to otopi
96974fa use dynamic inclusion to avoid tag inheritance
4a9a23f Force facts gathering on partial executions
7428b54 Use static imports and tags
9267118 Fix rhbz1654697 (#108)
5e928a1 build: post ovirt-ansible-hosted-engine-setup-1.0.8
8607d90 (tag: 1.0.8) build: ovirt-ansible-hosted-engine-setup-1.0.8


I see
10:59:42 2019-02-05 09:59:42,448:INFO:__main__:File 'automation/upstream_sources.yaml' cannot be opened
10:59:42 2019-02-05 09:59:42,449:INFO:__main__:Executing command: 'git --git-dir=/home/jenkins/workspace/oVirt_ovirt-ansible-hosted-engine-setup_standard-check-pr/ovirt-ansible-hosted-engine-setup/.git --work-tree=/home/jenkins/workspace/oVirt_ovirt-ansible-hosted-engine-setup_standard-check-pr/ovirt-ansible-hosted-engine-setup reset --hard'
10:59:42 2019-02-05 09:59:42,472:DEBUG:__main__:Git exited with status: 0
10:59:42 2019-02-05 09:59:42,472:DEBUG:__main__:  ---- stderr ----
10:59:42 2019-02-05 09:59:42,472:DEBUG:__main__:  ---- stdout ----
10:59:42 2019-02-05 09:59:42,472:DEBUG:__main__:    HEAD is now at af00824 Merge ceca0a852d94bd5af7226b15d15c9e3ff917cd5c into 9267118f456de46b9059e76b63bc85c18fcab5dd

Now, ceca0a852d94bd5af7226b15d15c9e3ff917cd5c is the last commit on my develop branch on https://github.com/tiraboschi/ovirt-ansible-hosted-engine-setup

commit ceca0a852d94bd5af7226b15d15c9e3ff917cd5c (origin/nolog)
Author: Simone Tiraboschi <stirabos@redhat.com>
Date:   Tue Feb 5 10:56:58 2019 +0100

    Avoid using no_log when we have to pass back values to otopi

which corresponds to b5a6c1d on master on https://github.com/oVirt/ovirt-ansible-hosted-engine-setup 

so "Avoid using no_log when we have to pass back values to otopi" got tested over 9267118 - Fix rhbz1654697 (#108) 
skipping
96974fa use dynamic inclusion to avoid tag inheritance
4a9a23f Force facts gathering on partial executions
7428b54 Use static imports and tags

in the middle.

The version bump change is in 7428b54
and so it builds as 1.0.9 and not as 1.0.10 as expected.


This is  a matter of timing - as I already said - I suspect the build you're looking at happened before b5a6c1d was merged.

In any case the build you're seeing ot the `check-pr` job has nothing to do with what gets sent into system testing and checked with OST. The build that gets checked is the one that is generated by the `on-ghpush` job and is always built from merged commits in the branch itself. You can see the last successful build is with the right version:


OK, and there we have ovirt-ansible-hosted-engine-setup-1.0.10-0.1.master.20190205155347.el7.noarch.rpm as expected.
Now the issue is why the CI job fails due to the lack of ovirt-ansible-hosted-engine-setup >= 1.0.10

Now I merged another patch to bump the version again to 1.0.11 just to re-trigger the whole flow.
Honestly I don't know what else I can do.


I don't think solving this was up to you. If the 1.0.11 build was removed from CQ because of an infra issue, Dafna or the infra-owner should have simply re-added it.

I did - twice

If it's a matter of repo ordering or whitelisting that causes an older version to be made available during is the run, it needs to be fixed in OST code.

It was available - that was no the issue.
since the issue was in upgrade only it may have been same version with diff hashing on our repos
For now we have a different issue which is blocking this one but if upping the version does not help I will delete packages in repo and re-add to CQ.

 


 
 
 

--
Barak Korren
RHV DevOps team , RHCE, RHCi
Red Hat EMEA
redhat.com | TRIED. TESTED. TRUSTED. | redhat.com/trusted


--
Barak Korren
RHV DevOps team , RHCE, RHCi
Red Hat EMEA
redhat.com | TRIED. TESTED. TRUSTED. | redhat.com/trusted


--
Barak Korren
RHV DevOps team , RHCE, RHCi
Red Hat EMEA
redhat.com | TRIED. TESTED. TRUSTED. | redhat.com/trusted