GitLab

looking at the logs is not good enough, makes much more sense to try a
sample call and wait until it succeeds.

add a puppet dependency for cci-tools repo, add policy-coreutils package
to get restorecon when using iptables-save.

replaces return which does nothing with an exit function, which calls
'exit' with the code on cmd failure.

until the new tellme_enable var change gets committed in
cloud_common/master.

disable the firewalld service (default in cc7) and fallback to iptables,
which our production modules rely on.

Fixes #1820.

mimic ciadm nodes, with an installation of all the openstack clients.
also installs the cci-tools and ai-tools, but for now these are not
really used (not sure we'll ever need them).

drop the openrc definition from site.pp as it comes from teigi (mimicing
production).

Fixes #OS-1532.

add a sample job config.

also updated instructions on how to rebuild the CI jenkins.

latest uses the 'latest' docker images of each OS container instead of
rebuilding from scratch. push pushes the current containers as the new
'latest' tag in the docker registry, to be done by CI only.

otherwise the puppet agent runs in the pods will fail to contact, as
puppetdb takes a few seconds to actually start serving requests.

also added some docs to the script.

jenkins will not have TTY set, and these flags are not needed.

plus the required dependencies, namely git-client and scm-api plugins.
also installs the git package in the master as the plugin relies on it.

for now we need to rely on the tbag_teigiurl teigi branch, until the
changes are merged to master.

a single script to prepare the workspace, launch the base cluster
including kubernetes, launch pods, etc.

this is the port we're using in the teigi container now.

default set to the sql driver.

the empty lemon class is a temporary workaround while we don't integrate
it into the dev environment. the osrepos include is required for a
couple hiera checks to work when doing osrepos::kojitag.

add instructions on how to recreate the heat stack holding the jenkins
master and the auto scaling for the slaves.

we need some post setup cmds like setting up the ceph pools. change the
check for pods ready from being only ceph to all pods.

add the required patch to get the kubernetes local cluster configured
with the dns server. also enables host networking, which comes disabled
in the kubelet by default.

we now run the ceph keyring creation from inside the container, so we
should reference /etc/ceph.

we're supposed to store the whole contents of the keyring file, not only
the value of the key (as we're now using puppet to configure this
instead of mounting the /etc/ceph mount in the container).

get hiera to populate the glance ceph.conf file, and make use of the new
flume agent hiera param to disable it in glance when needed.