Support of multiple MAC addresses for host declaration and a little fix to failover

[redrampage]

I'm back again with a new PR.

I've run into some problems with static DHCP reservations on bonded interfaces:

• The resulting MAC address assigned to the bond can be inherited from any slave interface, whichever is first added to the bond.
• When using iPXE to boot from LACP enabled interfaces it will choose any of the slave interfaces whichever discovered first.

So I wrote a small patch to support assignment of multiple MACs to a single host. It creates a separate host statements for each MAC(because dhcpd wants them to be unique).
If no 'option host-name' is specified in 'host.options' - it is added. This is made for compatibility with 'use-host-decl-names on' dhcpd option.

Also there is a small patch for failover configuration which allows usage of hostnames in inventory file.

And some fixes for group expanding macros.

[drybjed]

I don't think these changes should be here. print.group() is a recursive call of group() macro from dhcpd.conf.j2 template. Without this you cannot do recursive groups in the configuration file. The same goes for print.subnet() and print.hosts(). You should reverse those changes.

[redrampage]

Hm, actually, on the flat group it gave me an error 'print is undefined'.
Ok, I'll look into it more.

[drybjed]

Check if creating nested groups works in dhcpd.conf template, using inventory variables.

[redrampage]

Hm, to me it looks like both variants are failing.
I've tried the following empty playbook:

---
- hosts: all
  sudo: yes
  vars:
    dhpcd_probe: False
    dhcpd_groups:
      - comment: "level one - elem one"
        groups: 
          - comment: "level two - emen one"
            options: "#testing"
          - comment: "level two - emen two"
            options: "#testing-2"
      - comment: "level one - elem two"
  roles:
    - role: debops.dhcpd

With old 'print.' statements(original contents) it fails with message:

AnsibleUndefinedVariable: One or more undefined variables: 'print' is undefined

Without it(my flawed patch) on recursive groups it fails with:

AttributeError: 'dict' object has no attribute '__call__'

I guess it's because 'groups' is already defined as local variable in the for loop.
I don't see any solutions except to rename those macros to 'print_group', 'print_host', etc... as they used to be once.

[drybjed]

Renaming could be a solution. I'll check the issue in a day or two when I have a chance (holidays).

[redrampage]

I've reverted macros names back to long and distinct ones. Also switched DHCP Probe off by default as mentioned in #12 (left lib paths as is, because don't know which ones to use). And finally added default value for default subnet(otherwise task breaks, is there is no default IPv6 defined on host), didn't thought of anything better than loopback for that.

[drybjed]

Looks good. I wonder about that IPv6 default value though, how did you get the error. Currently debops.dhcpd is written to manage either DHCPv4 or DHCPv6 server, and it expects the respective network / IP addresses present and configured. In other words, if you don't have IPv6 enabled, how did you get an error from DHCPv6? Unless it's an issue related to debops.dhcpd and debops.subnetwork role order in the main playbook...

[redrampage]

When I run setup module on my test host

ansible -m setup -i ./inventory test.host

I get an empty dict in "ansible_default_ipv6", so the dhcpd role basically fails with error "address not defined in ansible_default_ipv6"

[drybjed]

Ah, during the Jinja templating phase then? Yeah, I suppose it would... OK, I'll accept ::1 as a default there, we'lll see if there are any unintended consequences.

Thanks for the contribution. :-) I'll update the documentation later, after I'm back from holiday.