cc48a5cd89
openSUSE 15.0 does not have libffi48-devel, instead we can use libffi-devel. Install libffi48-devel only on openSUSE 42.3. This was triggered by the failure in https://review.openstack.org/617282 Change-Id: I2207d69bd837a7249476b4a20025f41df3a7bc84 |
||
|---|---|---|
| .. | ||
| cache-devstack | ||
| infra-package-needs | ||
| initialize-urandom | ||
| nodepool-base | ||
| openstack-repos | ||
| zuul-worker | ||
| bindep-fallback.txt | ||
| README.rst | ||
Using diskimage-builder to build devstack-gate nodes
In addition to being able to just download and consume images that are the same as what run devstack-gate, it's easy to make your own for local dev or testing - or just for fun.
Install diskimage-builder
Install the dependencies:
sudo apt-get install kpartx qemu-utils curl python-yaml debootstrapInstall diskimage-builder:
sudo -H pip install diskimage-builderBuild an image
Building an image is simple, we have a script!
bash tools/build-image.shSee the script for environment variables to set distribution, etc. By default it builds an ubuntu-minimal based image. You should be left with a .qcow2 image file of your selected distribution.
Infra uses the -minimal build type for building Ubuntu/CentOS/Fedora. For example: ubuntu-minimal.
It is a good idea to set TMP_DIR to somewhere with
plenty of space to avoid the disappointment of a full-disk mid-way
through the script run.
While testing, consider exporting DIB_OFFLINE=true, to skip updating the cache.
Mounting the image
If you would like to examine the contents of the image, you can mount it on a loopback device using qemu-nbd.
sudo apt-get install qemu-utils
sudo modprobe nbd max_part=16
sudo mkdir -p /tmp/newimage
sudo qemu-nbd -c /dev/nbd1 /path/to/devstack-gate-precise.qcow2
sudo mount /dev/nbd1p1 /tmp/newimageor use the scripts
sudo apt-get install qemu-utils
sudo modprobe nbd max_part=16
sudo tools/mount-image.sh devstack-gate-precise.qcow2
sudo tools/umount-image.shOther things
It's a qcow2 image, so you can do tons of things with it. You can upload it to glance, you can boot it using kvm, and you can even copy it to a cloud server, replace the contents of the server with it and kexec the new kernel.