Jump to: navigation, search

Difference between revisions of "StarlingX/Tips Tricks Debugging Techniques"

(Resolution)
 
(4 intermediate revisions by 2 users not shown)
Line 75: Line 75:
  
 
[[File:StarlingX Hades Canyon Worker Set parameters.jpg|StarlingX Hades Canyon Worker Set parameters]]
 
[[File:StarlingX Hades Canyon Worker Set parameters.jpg|StarlingX Hades Canyon Worker Set parameters]]
 +
 +
 
<br />
 
<br />
  
 
=== Tips For Configuring NVMe Drive as Primary Disk  ===
 
=== Tips For Configuring NVMe Drive as Primary Disk  ===
 
[https://docs.starlingx.io/deploy_install_guides/nvme_config.html Nvme Tricks]
 
[https://docs.starlingx.io/deploy_install_guides/nvme_config.html Nvme Tricks]

Latest revision as of 19:10, 1 July 2020

Hardware specific Installation Guide

Intel Haydes Canyon NUC Issues and Resolution

Problem Description - Removing generic driver

New NUC8 (NUC8i7HVK, these are the dual NIC NUC's with a skull on them). When we switched from the 3.10.0-957 kernel to the 3.10.0-1062 kernel in master the system started to hang. The initial kernel boot was hanging once the initialization has started and the issue seemed like the Framebuffer initialization.

The following message was the last line:

FB: conflicting fb hw usage amdgpudrmfb vs EFI VGA - removing generic driver
Resolution
After adding "nomodeset" kernel cmdline parameter it seems to have solved the problem for AIO Simplex Installation.


Problem Description - Failed to identify rootdisk via pvdisplay

New NUC8 (NUC8i7HVK) with 2 1TB nvme as storage. When we try to Install Release 2.0/3.0/4.0 the installation fails with below error:

Storage Configuration failed: Requested boot drive "/dev/disk/by-path/pci-0000:72:00.0-nvme-1" doesn't exist or cannot be used

The boot drive does exist as manual Installation steps like setting up the "Installation Source" as /dev/nvme0 or /dev/nvme1 did start the installation process.

But Post-Installation Script failed with another error:

Error: Failed to identify rootdisk via pvdisplay

Resolution

Please replace the 1 TB nvme with 512GB nvme and installation process works fine.


Problem Description - System must have a ceph backend

New NUC8 (NUC8i7HVK) with 2 1TB nvme as storage. When we try to Add an OSD on controller-0 for Ceph using below command:

system host-disk-list controller-0 | awk '/\/dev\/sdb/{print $2}' | xargs -i system host-stor-add controller-0 {}

You might encounter below error:

System must have a ceph backend

Resolution

Configure Ceph Backend with below command

system storage-backend-add ceph --confirmed


Problem Description- Failing to boot up as worker node

New NUC8 (NUC8i7HVK) with 512 GB nvme SSD and 32 GB RAM. When we try to add it as the worker node ( tested in the Duplex Extended setup) in the controller Was same for lowlatency and normal subfunction case

$system host-update 3 personality=worker hostname=worker-2 subfunctions=lowlatency rootfs_device=/dev/nvme0n1 boot_device=/dev/nvme0n1

In the worker screen, we got personality as UEFI Lowlatency.

But you might encounter below error:

 blank screen with no progress showing of installation

Snapshot of the parameters that appear in worker console to install Stx was as below:

StarlingX Hades Canyon Worker Node kernel boot parameter

StarlingX Hades Canyon Worker Node kernel boot parameter


The issue could be because of dual graphics from Intel and AMD Radeon

Resolution

Configure Set parameters when the personality of worker node appears in the screen by pressing tab for editing kernel boot parameter by (see link below for better clarity):

Adding inst.graphical, nomodeset and removing console= ttyS0,115200, inst.gpt and inst.txt as shown in the image link below 

StarlingX Hades Canyon Worker Node kernel change boot parameter

StarlingX Hades Canyon Worker Set parameters



Tips For Configuring NVMe Drive as Primary Disk

Nvme Tricks