Even with the docker proxy cache we often get
jobs failed due to Docker Hub rate limits.
As per recommendation from the Opendev Infra team
let's pull as many as possible images from other
registires.
This PR updates the dnsmasq and nginx images used
for auxiliary purposes during deployments.
Change-Id: I58946e6fc63d726e08d83ea7f96e7fef140ddf21
Here we add Ansible tasks to the deploy-env role
to setup passwordless ssh from the primary node
to K8s cluster nodes. This is necessary for some
test scripts like for example Ceph migration script.
Change-Id: I1cae1777d51635a19406ea054f4d83972e5fe43c
Forward requests for unknown names to 8.8.8.8
NOTE: Temporarily disable DPDK job which turned to
be incompatible with this PR
https://review.opendev.org/c/openstack/openstack-helm/+/914399
It wasn't tested with the DPDK job.
Change-Id: I936fb1032a736f7b09ad50b749d37095cce4c392
- Make it less mixed. Each task file
deploys one feature.
- Deploy Metallb
- Deploy Openstack provider network gateway
Change-Id: I41f0353b286f817cb562b3bd59992e4baa473568
We are going to use containerized Openstack client
in test scripts. Adding zuul to the docker group
allows running docker command directly not using sudo.
Change-Id: Iee77e7f2b8801743f95535d31d0b909dcea50bf3
- In case we deploy Ceph on a multi-node env we have
to prepare the loop devices on all nodes. For this
we moved loop devices setup to the deploy-env
Ansible role.
For simplicity we need the same device on all nodes,
so we create a loop device with a big
minor number (/dev/loop100 by default) hoping
that only low minor numbers could be busy.
- For test jobs we don't need to use different devices
for OSD data and metadata. There is no
any benefit from this for the test environment.
So let's keep it simple and put both OSD data and metadata
on the same device.
- On multi-node env Ceph cluster needs cluster members
see each other, so let's use pod network CIDR.
Change-Id: I493b6c31d97ff2fc4992c6bb1994d0c73320cd7b
This PS mounts extra 80Gb volume if available and mounts it to
/opt/ext_vol. It also alters docker and containerd configs to move their
root folder to that extra volume. This helps zuul gates to succeed when
a node with 40Gb volume is assigned to a zuul gate.
Change-Id: I1c91b13c233bac5ebfe6e3cb16d4288df2c2fe80
This role works both for singlenode and multinode
inventories. The role installs all necessary prerequisites
and deploys K8s with Containerd as a container runtime.
The idea is to use this role to deploy
all test singlenode/multinode environments for all test jobs.
This PR wraps into a role playbooks that
we are currently using for multinode compute-kit tests.
Change-Id: I41bbe80d806e614a155e6775c4505a4d81a086e8
