[AutoQA] #245: Add support for test re-scheduling
by fedora-badges
#245: Add support for test re-scheduling
--------------------+-------------------------------------------------------
Reporter: kparal | Owner:
Type: task | Status: new
Priority: major | Milestone: 0.5.0
Component: core | Keywords:
--------------------+-------------------------------------------------------
After we start sending results to Bodhi/ResultDB there will immediately
emerge a need for test re-scheduling. There may be different reasons for
that:
1. Something went wrong in the test execution, we need to run it again.
2. The environment has changed and test needs to be executed once again.
For example the upgradepath test for some update failed because of invalid
package versions in higher Fedora releases. That problem has been
corrected and we need now a new execution of the test.
The important thing is that the maintainer should be able to trigger the
re-scheduling, not just us.
Find a way how to properly reschedule tests (we will probably utilize
ResultDB for that) and add that functionality.
--
Ticket URL: <https://fedorahosted.org/autoqa/ticket/245>
AutoQA <http://autoqa.fedorahosted.org>
Automated QA project
13 years
[AutoQA] #212: fix check_opt_in to not use collection_name
by fedora-badges
#212: fix check_opt_in to not use collection_name
-----------------------+----------------------------------------------------
Reporter: wwoods | Owner:
Type: task | Status: new
Priority: minor | Milestone:
Component: docs/wiki | Version: 1.0
Keywords: |
-----------------------+----------------------------------------------------
collection_name was used to line up with the dist-cvs branch layout; now
that we have switched to dist-git/fedpkg the branch names match the repo
names used in repoinfo, so we have no need for collection_name.
This will require modifying the existing autoqa-optin file structure to
match the git layout. We can leave symlinks for the old ('F-14') distro
names but use the new ones ('f14') as the canonical ones.
--
Ticket URL: <https://fedorahosted.org/autoqa/ticket/212>
AutoQA <http://autoqa.fedorahosted.org>
Automated QA project
13 years
[AutoQA] #247: upgradepath: possible broken-updates scenario for new packages
by fedora-badges
#247: upgradepath: possible broken-updates scenario for new packages
--------------------+-------------------------------------------------------
Reporter: kparal | Owner:
Type: defect | Status: new
Priority: major | Milestone:
Component: tests | Keywords:
--------------------+-------------------------------------------------------
Here is a scenario that will create broken upgrade path and our current
tests will not catch it:
1. Package foo is available for F12 (foo-1.0-1.fc12), but not for F13.
2. foo-1.0-1.fc13 is sent to f13-updates-pending as a new package update.
upgradepath passes.
3. foo-2.0-1.fc12 is sent to f12-updates-pending. Since there is still no
foo in F13 (AutoQA passed but it has not been pushed by RelEng team),
upgradepath passes.
4. foo-2.0-1.fc12 is pushed to f12-updates.
5. foo-1.0-1.fc13 is pushed to f13-updates.
6. Broken upgrade path.
What can we do about it?
We can break the scenario between steps 4 and 5. But we can't probably do
that just by watching -pending tags. This requires cooperation from the
updates manager (i.e. Bodhi). Bodhi must know that if some foo build
enters f12-updates, and there is also another foo build in f13-updates-
pending, it must notify us (over Fedora Message Bus) and ask us to retest
the latter build.
We can also break the scenario between steps 2 and 3. When we run
upgradepath on foo in f12-updates-pending, we can take into account not
just f13-updates, but also f13-updates-pending. But this is quite
controversial. Some builds in -updates-pending may be denied by RelEng
team, but we could have failed some other builds because of them.
We will probably solve this ticket only when we have advanced
infrastructure ready (ResultDB, Fedora Message Bus). Without it this seems
pretty hard to be solved properly.
--
Ticket URL: <https://fedorahosted.org/autoqa/ticket/247>
AutoQA <http://autoqa.fedorahosted.org>
Automated QA project
13 years
[AutoQA] #246: Add temporary hack for test re-scheduling
by fedora-badges
#246: Add temporary hack for test re-scheduling
--------------------+-------------------------------------------------------
Reporter: kparal | Owner:
Type: task | Status: new
Priority: major | Milestone: 0.4.4
Component: core | Keywords:
--------------------+-------------------------------------------------------
Until we have a proper support for test re-scheduling (ticket #245) we
need some temporary hack to enable test re-scheduling in most common
cases.
We plan to have upgradepath and depcheck sending comments to Bodhi. There
is probably no problem with depcheck, because it re-evaluates the whole
contents of -pending tag on its every update. But there are some issues
with upgradepath, one use case is described in ticket #245.
We need to:
1. get a list of use cases for which we need to reschedule upgradepath or
depcheck
2. inquire whether Bodhi changes timestamp (or whatever) of the updates
in those cases (i.e. we would detect it and run tests again properly even
with our current watcher implementation)
3. in the remaining cases decide what to do with it. For example we can
make temporary upgradepath behave the same way as depcheck does, i.e. test
all builds in -updates-pending for every update of that tag. Because we
will have Bodhi comment duplication prevention, it could work quite fine.
--
Ticket URL: <https://fedorahosted.org/autoqa/ticket/246>
AutoQA <http://autoqa.fedorahosted.org>
Automated QA project
13 years, 1 month