The httpd wasn't able to open a new socket for connecting to identity server,
this fix will allow that.
Signed-off-by: Michal Konečný <mkonecny@redhat.com>
This kinda goes along with:
https://github.com/fedora-infra/bodhi/pull/4759
but neatly, we don't need to wait for that to be merged or
deployed, we can do it right away and it'll work fine now but
also DTRT when we switch to grouped mode in Bodhi. This splits the
greenwave policies for openQA update tests into groups according
to the test flavors, and has each grouped policy apply to what
will be the correct decision contexts for Bodhi's grouped critpath
support. We *also* have each grouped policy apply to the existing
non-grouped contexts, so as long as Bodhi is running in
non-grouped mode, things should behave exactly the same: all
critpath updates will be gated on all the tests.
We ensure every critpath group is listed in the null policies,
to make sure we don't get errors when Bodhi queries against them.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
300G doesn't seem like enough when we have candidate composes
alongside Rawhide and Branched, we're getting incomplete tests
because assets are being removed.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
Prevent these jobs from running so long that new jobs can get started
meanwhile. batcave has been observed to run sometimes dozens of old
list-vms-per-host processes under unusual but not rare circumstances.
This commit changes three Fedora mail aliases as the FCAIC role
transitions from @riecatnor to @jflory7:
* `fcaic`: Changed from `riecatnor` to `jwf@`
* `matrixadmin`: Changed from `mattdm` and `riecatnor` to `fpl@` and
`fcaic@`
* `podcast`: Changed from `riecatnor` to `fcaic@`
Note that I did not use my FAS username but my mail alias for the
`fcaic` address. I'm not sure if it is required to be a FAS account or
if an alias is acceptable.
CC: @mattdm@bcotton
Signed-off-by: Justin W. Flory (he/him) [Red Hat] <jwf@redhat.com>
This reverts commit 892453da7e.
openQA still had problems with the very long request, so I just
did an ugly hack to get the request under the limit instead.
The openQA job scheduler was hitting 414 errors today because
an update has so many builds there are more than 8190 characters
(the default limit) in the POST request. Let's bump the limit
to 16000.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
Right now there's often a backlog due to it not responding to the fedora
side. So, lets bump up the checks a bit so they do not alert all the
time.
Signed-off-by: Kevin Fenzi <kevin@scrye.com>
If you run these playbooks without any limit, The 4 master hosts (2
x86_64 and 2 aarch64) will be in the play, but they all use local_action
to make the local secrets file thats loaded. This means, whichever of
them happens to be writing the file last, thats the version of the file
that all 4 of them get. This is particularly bad when it's the staging
creds and the prod hosts get it loaded. :(
So, adding {{ env }} here makes the staging and prod versions seperate
so they don't step on each other.
Signed-off-by: Kevin Fenzi <kevin@scrye.com>
With the release of 1.6.0 we can now remove the poetry specific changes for
staging and instead use the same for staging and production.
Signed-off-by: Michal Konečný <mkonecny@redhat.com>
