The for loop in Groovy does not work, but fortunately it seems that the
CentOS CI network environment is stable again. Pulling container images
is working as it did before.
Signed-off-by: Niels de Vos <ndevos@ibm.com>
Jenkins Jobs Builder does not like the inline `dsl:` scripts anymore, so
replace them by Groovy files in the main directory, similar to other
jobs.
Signed-off-by: Niels de Vos <ndevos@ibm.com>
The `RUN_DISPLAY_URL` is automatically set by Jenkins, but it seems the
latest Jenkins Jobs Builder update does not detect it automatically.
Run the JJB commands with `--allow-empty-variables`.
Signed-off-by: Niels de Vos <ndevos@ibm.com>
The output of the for loop seemt to be `groovy.lang.IntRange`, which is
not very useful. Maybe it is required to define the variable of type
`int`?
Signed-off-by: Niels de Vos <ndevos@ibm.com>
There seems to be some failing network component in the CI environment
that prevents pulling container images. Even pulling it from the
registry to my laptop fails often, but with a few retries everything
gets pulled eventually.
Signed-off-by: Niels de Vos <ndevos@ibm.com>
The test_type and deployment related args are being passed
as make option instead of the E2E_ARGS. Pass it as E2E_ARGS.
Signed-off-by: karthik-us <ksubrahm@redhat.com>
The test_type is being passed as the make option instead
of the E2E_ARGS. This fix should pass it as the E2E_ARGS.
Signed-off-by: karthik-us <ksubrahm@redhat.com>
The current e2e test runs will trigger all 3 types of
tests which takes a lot of time to complete. If the
patch impacta only a single type, then having an option
to trigger only that particular e2e test can save
considerable amount of time based on the test type.
This can also help in reducing the time in testing when
a patch failed on a particular e2e test type and triggering
only that e2e with the fix for the failure or re-trigger
only that test again to check for false positives.
Signed-off-by: karthik-us <ksubrahm@redhat.com>
few commands are getting permission denied error
adding sudo is fixing the problem
```
Warning: Permanently added '192.168.49.2' (ECDSA) to the
list of known hosts.
Hint: You are currently not seeing messages from other
users and the system.
Users in the 'systemd-journal' group can see all messages.
Pass -q to turn off this notice.
No journal files were opened due to insufficient permissions.
```
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
CI jobs pull the ceph/ceph:v18 image (without `quay.io`). If the
mirroring includes the registry host, the image can not be found.
Updates: #4030
Signed-off-by: Niels de Vos <ndevos@ibm.com>
As we need to move to latest Rook
and Ceph releases in devel branch
updating mirroring images to push
latest images to internal registry.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
This commit enables overriding csi_upgrade_version
used for upgrade testing with CSI_UPGRADE_VERSION
value from build.env.
Signed-off-by: Rakshith R <rar@redhat.com>
The `/test ci/centos/mini-e2e-helm` has been broken with the last change
where `/test all` was removed. A remaining '(' makes the regular
expression invalid, and jobs fail to get started.
Fixes: f9f3bb9 ("ci: remove `/retest all` command for Jenkins jobs")
Signed-off-by: Niels de Vos <ndevos@ibm.com>
`/retest all` causes a spike in resource consumption in Jenkins and the
OpenShift cluster kills the Pod. That means tests are not fully running
yet, and results never arrive back in the PR. Instead of `/retest all`,
the `ok-to-test` label can be used to trigger required tests with a
slight delay between each command.
Signed-off-by: Niels de Vos <ndevos@ibm.com>
The upstream reported issue suggests adding a `timeout` value in the
Jenkins Jobs Builder configuration. This looks like a nicer workaround
than downgrading urllib3.
Fixes: #3946
See-also: https://storyboard.openstack.org/#!/story/2010752
Signed-off-by: Niels de Vos <ndevos@ibm.com>
jenkins-job-builder does not seem to work with urllib3 2.x, force using
version 1.x as a workaround.
Updates: #3946
Signed-off-by: Niels de Vos <ndevos@ibm.com>
K8s 1.24 will be End of Life on 2023-07-28.
Therefore, this commit removes support
for k8s 1.24.
refer:
https://kubernetes.io/releases/#release-v1-24
Signed-off-by: Rakshith R <rar@redhat.com>
CentOS Stream 8 does not provide a module named 'dataclasses'. It seems
current versions of Jenkins Jobs Builder require that. Python 3.7
offers the module by default, so using a new base distribution with
Python 3.9 has the module too.
See-also: #3909
Signed-off-by: Niels de Vos <ndevos@ibm.com>
removed ability to run tests with
kubernetes 1.23 as its no longer
maintained version.
fixes: #3636
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
In the old OpenShift cluster the registry consumer 800+ GB of data. Once
running the garbage collection manually, the consumption reduced to a
little over 8GB. Let's be nice users of the infrastructure and run
garbage collection weekly.
Signed-off-by: Niels de Vos <ndevos@ibm.com>
ShellCheck failed with the following warning:
```
SC2086 (info): Double quote to prevent globbing and word splitting.
Did you mean:
git config --global --add safe.directory "${PWD}"
```
Signed-off-by: Niels de Vos <ndevos@ibm.com>
When Jenkins Jobs have been modified, they should get deployed in the
Jenkins environment. This seems to fail with the following error:
```
Initialized empty Git repository in /opt/build/.git/
fatal: detected dubious ownership in repository at '/opt/build'
To add an exception for this directory, call:
git config --global --add safe.directory /opt/build
```
By marking /opt/build as a safe directory in the global git
configuration file, this problem should not occur anymore.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
It seems all jobs were converted to the new `duffy` command, except
ci-job-validation which still used the old `cico` command.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
The `cico` command is getting deprecated, and the new `duffy` command
should be used instead. The new command requires a `~/.config/duffy`
file that contains an API endpoint and credentials.
See-also: https://sigs.centos.org/guide/ci/#installing-and-configuring-duffy-client
Signed-off-by: Niels de Vos <ndevos@redhat.com>
Upgrade tests always runs with the latest
release of kubernetes, updating the
kubernetes to latest available release
in the CI.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Disable e2e tests by default in CI to
save resources. cephcsi maintainers or
contributors can add the ok-to-test label
to trigger the e2e tests.
depends on #3468
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
Installing Helm fails if `wget` is not available. The script that
installs Helm does not seem to abort though.
By installing `wget` in preparation for deploying with Helm, the CI jobs
should get ready to run successfully.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
Now that we run with Podman, the journal from the container that was
started by minikube is not very useful. In addition to the journal from
the minikube container or VM, include the journal from the host.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
Even when minikube is running with `--driver=podman`, `minikube ssh`
works and container images can be copied into the minikube container.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
With the CentOS CI machines moving to AWS EC2 virtual-machines, there is
no option to run minikube with a VM anymore. Instead, run minikube with
the Podman driver and partition the extra xvdb EBS volume into three
pieces.
Signed-off-by: Niels de Vos <ndevos@redhat.com>
