Focus on supporting proxy. (But (un)commenting a few lines and you get non-proxied access)
All images have:
proxy support except when the name ends with "-np" (no proxy).
VirtualBox Guest Additions
Vagrant export as a bonus
vb-fedora-27 (VirtualBox from Oracle repo) Fedora 27
The point is to script the build.
I will try to add:
packerio! Installed in /usr/local/bin to be in the PATH)proxy-cleaner.sh.ssh-keygen -t ecdsa or enable passwords) in the Packer file (.json) and the kickstart file, I also add the Vagrant user and key.Could help: sudo dnf install pykickstart ansible squid
I deal with 2 proxy:
If you don't need them, it is easy to change the kickstart file to link to non-proxied sources.
I configure the Guest to use the Corporate proxy to access Internet.
See the script proxy-cleaner.sh
The proxy will be configured for DNF and as an environment variable. This is done in a shell provisioner.
Also do not forget to configure the local RPM cache (/etc/squid/squid.conf) to use the Corporate proxy.
To avoid repetitive downloads of the same RPM, I setup a local cache with Squid and Apache on my local laptop.
Once installed (see below), you can launch it with rpm-cache/start.sh I did not make it permanent, because I do not want to have Squid and Apache and security settings always on.
Your kickstart file must then use http://10.0.2.2/ or http://192.168.56.1/ depending on the interface setup in the packer template.
Inspiration:
An install script would looks like this (you may want to manually edit your squid config file):
sudo dnf install squid httpd
sudo cat >> /etc/squid/squid.conf <<EOF
# Uncomment the following 2 lines to use a Corporate proxy (see http://wiki.squid-cache.org/Features/CacheHierarchy)
#cache_peer Proxy.MyCorporate.net parent 80 0 no-query default login=YourLogin
#never_direct allow all
cache_replacement_policy heap LFUDA
maximum_object_size 8192 MB
cache_dir aufs /var/spool/squid 16000 16 256 max-size=8589934592
acl repomd url_regex /repomd\.xml$
cache deny repomd
EOF
sudo cp rpm-cache/dnfcache.conf /etc/httpd/conf.d/
# You have to explicitly create the Squid folder or SELinux will not allow the Squid service to write them (Squid won't even start without)
sudo squid -zSimply run ./make.sh filexxx<.json>. (It can take hours.)
And then ./launch.sh filexxx<.json>
The proxy should not be asking for a password. But in any case: Be careful not to commit your proxy password in the repository!
SELinux is enabled in the Guest. (There is one line to uncomment in the kickstart file to disable it)
Check the SSH key in kickstart (public) and Packer json template (path to private)
You may addi root password in the kickstart for debugging.
I setup 2 network interfaces:
Once the VirtualBox extension are installed, one can ask for the "Host-only" IP with VBoxManage tools and SSH to the box. (see the launch.sh script)
Beware that the VirtualBox virtio driver has good perfomance, but CentOS is not capable of using enp0s3 names with it. CentOs will fallback on eth0 names!! See https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/html/Networking_Guide/sec_Troubleshooting_Network_Device_Naming.html
Usually, the default install is OK. You can let autopart in kickstart do its job.
But you may need a particular setup. Typically, OpenShift master (in my example) requires to have free space in the Volume Group (LVM). In this case, the kickstart contains specific instruction.
If you need 2 disks, it is difficult to automate reliably! Instructions are specific to VirtualBox (Packer does not help you here) and if you want to recover errors, you may need particular scripts (outside Packer). I decided to stop te complexity to just creating the second disk. In case of errors, you have to use the VirtualBox GUI "File/Virtual Media Manager". Or on the command line you could do
vboxmanage list hdds
VBoxManage storageattach <uuid|vmname> --storagectl <name> --device <number> --medium none # Eventually detach the disk
vboxmanage closemedium disk <uuid|filename> --delete Also note that the second disk will be created in the current folder as VDI, then exported in VirtualBox as VMDK.
OpenShift Origin : There is a Master machine and a Node machine.
They each use different disk setup to illustrate the possibilities.
They use a static IP, so the VirtualBox addition could be avoided (it would speed up install and reduce the size). Actually, the kickstart file is almost suffisient.
I tried to reduce the size of the Docker-Storage-Setup. But it is hard to go beyond 8Go. Maybe we need to play with parameters (DATA_SIZE, DATA_MINIMUM_SIZE ...)
Maybe I go too far with the kickstart file and an ansible or shell script would be more the right tool for the job. But currently, nothing too complex is done.
To install OpenShift, you still have to distribute the key:
ssh-keygen -f /root/.ssh/id_rsa -t rsa -N ''
for i in 192.168.56.10 192.168.56.12;do ssh-copy-id root@$i;doneand this will require a password for root (look at the kickstart file). It should be possible to installl from the linux host machine so (since you already have an account ssh key. I have to test if it works !!!
For usual stuff. There is a Vagrant export for Fedora.
On Linux I had to specify the provider on the command line:
vagrant box up --provider virtualbox
OpenStack all-in-one (with RDO Packstack) (The OpenStack install fails if there is a proxy.)
OpenStack is huge.
Useful command lines available: https://www.virtualbox.org/manual/ch08.html
We could create a shared folder to ease exchanging files with the host. See files make.sh and filexxx.json.
On the host: /mnt/guest_vm and on the guest: /media/host_machine.