experience with multiple postinstall scripts?

[Julian Dunn]

Does anyone have experience with multiple postinstall scripts?

I tried to set this up in my definition but it didn't work. At the end of the first one, I reboot the VM, and I can see VirtualBox successfully reboots it, but Veewee doesn't seem to ever re-log-in and execute the second. It just hangs.

- Julian

[Patrick Debois]

On 20/12/12 00:37, Julian Dunn wrote:
> Does anyone have experience with multiple postinstall scripts?

Hi Julian,

multiple postinstall scripts are in use in several templates and they
work as supposed.

https://github.com/jedi4ever/veewee/blob/master/templates/CentOS-6.3-i386-minimal/definition.rb

> I tried to set this up in my definition but it didn't work. At the end
> of the first one, I reboot the VM, and I can see VirtualBox
> successfully reboots it, but Veewee doesn't seem to ever re-log-in and
> execute the second. It just hangs.

I guess the main difference is that you reboot during on of the
postinstalls.

Could you provide me with your definition, so we can check the behavior?

thanks you

[Julian Dunn]

Thanks Patrick. Here is what I was trying to do (ignore the rest of the diff, you've seen that before in my pull request)

https://github.com/juliandunn/veewee/commit/556609989af246b7d9f5eaa2f05b26ea14ff3168

The reason I'm trying to do this is because a "yum update" in the old postinstall.sh will update the kernel, but VirtualBox's kernel extensions get built against the old kernel.

- Julian

[Julian Dunn]

Okay. So Patrick, after taking your advice about putting a "service sshd stop && reboot" at the end of the first postinstall script, I came across other interesting behaviour. This is what happens:

https://gist.github.com/4596902

To work around this, I had to rescue for IOError on the outside of the block and move on to the next script, but I recognize this is a pretty awful solution:

https://github.com/juliandunn/veewee/commit/db5022d7723090e08926c486919016892e44b8dc

Thoughts on how I could do this better? It does work though, and I get a box built at the end.

- Julian

[Patrick Debois]

Hey Julian,

I actually think you're hack is a good solution - I'd be happy to merge it in.

Patrick

--
 
 

[Julian Dunn]

I'll put in a pull request to veewee. But I really think it's brittle at the moment. Net::SSH should just trap IOError and raise Net::SSH::Disconnect, that way it'd be cleaner and Veewee could trap that outside of the loop as a "normal" condition. I'll file a bug with the Net::SSH folks.

- Julian