I have a node that has successfully been commissioned into a MAAS pool, i.e., got Declared->Commissioned->Ready. After it got to Ready state, I issued juju add-machine to put it in Allocated state; however, upon boot up, I get the following messages:
....
cloud-init-nonet [18.88] waiting 10 seconds network device
cloud-init-nonet [20.93] waiting 120 seconds network device
cloud-init-nonet [148.93] gave up waiting for network device
...
This node is a Y500 Lenovo laptop, and I had to make the following modifications:
- Its BIOS is set to Legacy with boot priority set to UEFI 1st
- On the boot up grub menu, I added nomodeset, to the end of linux argument line.
It should be noted that the 2nd step was initiated, after node was set into Allocated mode, to avoid hang up on nouveau driver, during boot up. I also tried adding
ether=0,0,eth0
to the linux argument line to no avail.
My initial inclination to blame UEFI as the root cause of the problem, has proven to be unfounded. Here is what I see now:
- Intermittently, allocated nodes may get stuck waiting for network device, e.g., after 4-5 reboots my Y500 node finally got through, and I was able to deploy it as a nova-compute. I have also noted that rebooting an existing node, say cinder or another nova-compute, results in the same behaviour, which is rather odd, as 2 weeks ago I deployed them with no such problem.
- The work around, for deployed nodes, is to login in single mode, add a password for ubuntu user, reboot, login with new credentials, and issue
dhclient br0
. I've also have noted that prior to issuing dhclient, br0 is not listed in ifconfig. - The problem that I was observing with regard to MAAS server also wanting to obtain a dhcp IP address, seems to have been due to mismatch between
/etc/network/interface
and Network Manager. Setting the relevant fields identically in both, gets rid of the annoying dhclient process.
I haven't been able to ascertain any new information by going through cloud-inits log files. If there are suggestion, kindly, let me know.
(Not sure if this is relevant, but I started seeing this problem as I tried to add 2 more nodes, to existing 10 node setup.)