[1.9] Cluster downloads boot-images from managed network (pxe) instead of network used to connect to Region

[Nicolas Thomas]

To reproduce the bug :

 Have 1 Region controller with 2 network interfaces.

Connect a cluster controller to the second interface
setup PXE/DHCP on first of your RC.

Then when importing boot image from the CC it tries to connect to the PXE network of the RC and not the second interface (the one configured to connect to RC).

See drawing attached for better understanding.

[Andres Rodriguez (andreserl)]

Nicolas,

Could you please attach:

/var/log/*.log (maas.log, regiond.log, clusterd.log) from the region and cluster respectively.

Also:

/etc/maas/regiond.conf and /etc/maas/clusterd.conf (the latter from the cluster).

[Nicolas Thomas (thomnico)]

To reproduce the bug (networks are provided as examples) :

Have a RC+CC with eth0 on 172.16.0.1 configure to be PXE/DHCP net.
Have eth1 on 10.1.1.1

Create a CC with 2 interface : eth0 on 192.168.0.1 to manage pxe/dhcp
eth1 on 10.1.1.2 to rely to RC+CC.

Be sure CC can not route to RC+CC eth0 .. at all (wall)

Then try to import-images from CC only .. expectation is downloading from 10.1.1.1 , netstat shows that CC try to connect to 172.16.0.1 which is blocked.

Hope this helps,

[Blake Rouse (blake-rouse)]

I have this branch here that is a simple change that I believe will solve your issue.

https://code.launchpad.net/~blake-rouse/maas/fix-1514883/+merge/277201

You could easily patch this file on each cluster to see if this resolves your issue.