Cinder Image Fails on SuperMicro 2U Storage Server

[Rob Neff]

This fails on all of 4 of our 2U SuperMicro storage nodes.

Attachment: Kickstart file.

Software:
Fuel 5.1

OS: CentOS

Architecture:
1 1U SM Controller
9 1U SM Compute (Install fine)
4 2U SM Storage (Compute Image installs fine, Cinder image has bug below)

Hardware
We have 4 2U SuperMicro nodes with
- 2 8c Ivy Bridge Server
- 96 GB RAM
- 2 rear Hot Swappable OS drives
- 3 SuperMicro 3008 LSI Cards with latest IT Firmware and BIOS
- 24 Front Drives - 22 10k drives + 3 Intel SSD's

Repro Steps:
1. Select Node and assign Cinder Role with LVM
2. Assign Base System to either be on either Rear Hot-Swappable Drives or the Front SSD (it doesn't seem to matter which drive we choose)
3. Click Apply Changes to start Deployment

Result:

Machine reboots and tries to run Kickstart, pretty soon an error screen appears.

ERROR SCREEN:

Disabling protocol version 1. Could not load host key.

------------| Error Parsing Kickstart Config |----------------------------------

The following error was found while parsing the kickstart configuration file:

The following problem occurred on line 2 of the kickstart file:

Specified nonexistent partition 3 in partition command.

                                  [Reboot button]

[Oleksiy Molchanov (omolchanov)]

Can be related to https://bugs.launchpad.net/fuel/+bug/1291692

[Vladimir Kuklin (vkuklin)]

Rob, thank you for posting this bug. Could you please add screenshots of failed anaconda process ?

[Vladimir Kuklin (vkuklin)]

IMO, root cause may be related to the fact that disks were initializing too slowly, that's why kickstart could not find them. So the question is, how much time did it take between the node being discovered and sent to deployment?

[Vladimir Kuklin (vkuklin)]

Also, could you try following steps:

dockerctl shell cobbler

find file /var/lib/cobbler/kickstart/centos<whatever I do not remember full name>.ks

open it and add following string:

sleep 120

somewhere before partitioning starts, e.g. right after `zerombr` line

[Vladimir Kozhukalov (kozhukalov)]

Please run ls -lR /dev command and provide us with its output so as we are able to compare it with what is written in your kickstart. And if it is possible please attach diagnostic snapshot to this bug.

[Satish Salagame (satish-salagame)]

Here is the screen shot of the Error parsing kickstart config

[Stanislav Makar (smakar)]

Could you please provide diagnostic snapshot ?

[Rob Neff (rob-neff)]

To be clear, when you say diagnostic snapshot, do you mean a snapshot of the Fuel Master Virtual Machine or something else?

A snapshot of Fuel will be 4-5GBs. Is that what you want?

Thanks,
Rob

[Mike Scherbakov (mihgen)]

Diagnostic snapshot is tarball with logs. It's available via Support page on UI, there is a button to generate one.