Most servers fail to deploy UC via MAAS
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
snapd |
Fix Released
|
High
|
Ian Johnson |
Bug Description
I've tried following instructions here for setting up MAAS to deploy Ubuntu Core:
https:/
I've followed these instructions for both Ubuntu Core 18 and Ubuntu Core 20.
With Ubuntu Core 18, I was able to successfully deploy two machines in the 18 Tremont St. certification lab:
* jehan (Quanta D52B-1U with a plug-in NVMe device)
* ashbrook (an Orange Box v3 prototype using ASRock MT-C224
motherboards)
Several other servers did not deploy:
* cloaker (Lenovo SR665)
* aitken (Lenovo SR650)
* feebas (Cisco UCS C220)
* jolteon (Lenovo x3550 M5)
* mouchez (Lenovo SR670)
The MAAS "Logs" tab showed what appeared to be a successful deployment; but when the nodes rebooted, they showed the following error on their consoles:
findfs: unable to resolve 'LABEL=writable'
On the successful deployments to jehan and ashbrook, the root (/) filesystem has a label of "writable", which suggests a missing disk device driver or some other disk-related problem with the initrd written by the MAAS deployment ephemeral. I believe that all of the failed systems have hardware RAID controllers. So does jehan; however, jehan also has a plug-in NVMe device, and the successful deployment was to that device. The RAID disks in jehan were accessible (albeit unconfigured) once the system booted.
With Ubuntu Core 20, I was unable to deploy to any system until I installed the "dangerous" image. That image deployed to ashbrook, but not to any other node; even jehan failed. Jehan seemed to deploy successfully, according to its logs, but failed to come back up. I didn't see the "findfs" warning noted earlier, but it's possible I missed it.
Confirmed, I also attempted this in our other lab on different servers and got a 100% fail rate there.