DHCP configuration failure on 2 NIC deployment

Asked by Gowri

Hi ,

I used Fuel installer and Grizzly on Ubuntu.

-Fuel master installation is complete.

I used "HA + Nova-network FlatDHCP manager" configuration on 2 NIC deployment using the instructions and the switch configuration provided in link : http://docs.mirantis.com/fuel/fuel-4.0/pre-install-guide.html#untagged-ports

 I assigned the roles to nodes and provided network configuration . when i deploy, I am getting an error saying "DHCP Auto configuration Failed". Does it mean it is not able to detect the DHCP server through any of the interfaces ?

Initially , the nodes PXE boot using eth0 interface but at the stage of installing OS, it is showing the above mentioned error.

My doubt is :

1. Does it need IP address from DHCP of master on both the interfaces for all slave nodes ?

2. I tried to configure network interface manually in the slave nodes, I got an error " failed to retrieve preconfiguration file from http:// 10.20.x.x/cblr/svc/op/ks/system/node -x". I fail to understand what it means.

3. When the slave nodes PXE boot using eth0, it should continue with the OS installation right ? The fuel master provides the ubuntu boot image to the slaves and the slave machines already have an IP on eth0 ,therefore it should continue with OS installation .If that is correct , why does it need a new IP ?
(Please correct me if I am wrong in my understanding).

4.The cobbler generated dnsmasq.conf in the fuel master has dhcp on only one interface eth0.

Please help me out . I am quite new to this and I am finding it difficult to figure out what is wrong.
Is there anything particular to 2 NIC deployment that I am supposed to follow?Please help.

Thanks,

Question information

Language:
English Edit question
Status:
Solved
For:
Fuel for OpenStack Edit question
Assignee:
No assignee Edit question
Solved by:
Gowri
Solved:
Last query:
Last reply:
Revision history for this message
Nikolay Korshenin (nkorshenin) said :
#1

> I assigned the roles to nodes and provided network configuration . when i deploy,
> I am getting an error saying "DHCP Auto configuration Failed".
> Does it mean it is not able to detect the DHCP server through any of the interfaces ?

Could you clarify where did you see this message and at what time?
Screenshots would be very desirable.

Could you also generate and give us diagnostic snapshot from Fuel ( Home ==> Support ==> Generate Diagnostic Snapshot ).

Revision history for this message
Gowri (gouribudal) said :
#2

Hi,

Thanks for the response. As I mentioned , the slave nodes PXE boot using eth0 interface but at the stage of installing OS, it is showing the error "DHCP Auto configuration Failed".

The sequence of steps I have done so far is as follows:

--> Fuel master installation is successful.

--> In the Fuel dashboard the slave nodes are identified and I assigned roles to the slave nodes.

--> All the network settings are done in the Fuel dashboard and network is verified successfully.

--> I have assigned public/floating and private network on interface eth1, admin ,management and storage networks on interface eth0.

--> After I click on "Deploy changes " ,the slave nodes PXE boot and is followed by OS installation(I am trying with Grizzly on ubuntu ).

--> At the initial stage of OS installation , the slave nodes are displaying the error "DHCP Auto configuration Failed".

--> I tried to skip this auto configuration step and tried to configure network manually but that again resulted in error "failed to retrieve preconfiguration file from http:// 10.20.x.x/cblr/svc/op/ks/system/node -x".

 Please help me in understanding what might have gone wrong.I have some doubts in basic concepts which I have posted above , Can you please clarify so that I can understand better.

Thanks

Revision history for this message
Gowri (gouribudal) said :
#5

 Hi,

 I was trying to solve this error and I came across this link which says that " In a multiple NIC environment , server can cause the installer to try to use a different network interface from the one that was used by PXE "( link : https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Linux/6/html/Installation_Guide/sn-booting-from-pxe-x86.html ). Can you please help me out with this ?

Is this the issue that is causing problem ?

I tried editing the config files in /var/lib/tftpboot/pxelinux.cfg in Fuel master but I am still facing same problem.

NOTE : Once the slave nodes encounter the error, I tried to check the list of interfaces on slave nodes and I found that the interface which was used for pxe boot is now recognised as eth1 and viceversa.
i.e The slaves used eth0 for PXEboot but after we deploy changes and OS installation starts,the interface eth0 is now showing as eth1.

Revision history for this message
Miroslav Anashkin (manashkin) said :
#6

--> I have assigned public/floating and private network on interface eth1, admin ,management and storage networks on interface eth0.

Are you use untagged Admin network and VLAN tagged management and storage networks on interface eth0?
Untagged Admin network is mandatory.

Revision history for this message
Gowri (gouribudal) said :
#7

Hi,

Yes I am using Untagged admin.
 Storage and management network as tagged.

Revision history for this message
Syed (smghayur) said :
#8

I am having similar issue with the same error with Fuel. Was it resolved ?

Essentially, the Fuel Master is a DHCP server on admin network but what about Management Network ? Do I need the management to be on the same ethernet port as Admin network with VLAN-tag ?

My guess is that Admin network should be responsible for doing the OS install ?

Please help ?

Revision history for this message
Gowri (gouribudal) said :
#9

Hi,

Yes, I have admin network and management network on eth0.
Yes I think Admin network should only take care of OS installation. I followed the configuration as per link :http://docs.mirantis.com/fuel/fuel-4.0/pre-install-guide.html#example-1-ha-nova-network-flatdhcp-manager .

Are you trying on ubuntu or centos?

Revision history for this message
Miroslav Anashkin (manashkin) said :
#10

Greetings Syed,

Could you please explain, what is your Fuel version?
Fuel 3.x and 4.x are different.
4.x has more strict rules for admin network interface.

Kind regards,
Miroslav

Revision history for this message
Syed (smghayur) said :
#11

4.x

Finally, I was able to make it work. For some reason, Ubuntu was not getting deployed. CentOS went through fine.

Thanks,
Syed

Revision history for this message
Gowri (gouribudal) said :
#12

Hi Syed,

Ubuntu has a bug related to 2 Nic deployment where manually we have to force the use of same NIC used for PXE to be used again during OS installation. This bug is fixed in later versions of Ubuntu ,but we were facing this issue with new versions of Ubuntu as well.

This problem is not there in Centos.

Refer the following for details :

> The link says that " In a multiple NIC environment , server can cause the installer to try to use a different network interface from the one that was used by PXE "( link : https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Linux/6/html/Installation_Guide/sn-booting-from-pxe-x86.html ).

For Ubuntu , I made changes in the file pxelinux.cfg and also changed the eth1,eth0 to MAC addresses of eth0 and reboot ,then it proceeded with installation.

Thanks and regards,
Gowri

Revision history for this message
Syed (smghayur) said :
#13

Thanks Gowri for the detailed explanation. It took me a while before I gave up on Ubuntu and went with CentOS :). I will go back and try out Ubuntu with the workaround.

Thanks,
Syed