jlehtone wrote: ↑2021/12/25 12:05:28
alpha754293 wrote: ↑2021/12/24 18:58:31
I can't even get it to install after writing the 7.9.2009 ISO onto a USB drive using Rufus 3.17 (which is the latest version as of this writing).
Does that thumb drive function properly on other machines?
If you have a kernel that would boot if there were no IB card, then installation without the card would be logical.
One blacklist the offending modules with the kernel command-line options?
Once system is up, running, "better" drivers are at hand and load during runtime, then make them load on boot too.
Yes, it does.
I am able to use the same thumb drive to perform the installation on an Intel Core i7-3930K on an Asus X79 Sabertooth motherboard and that system also has an IB card as well.
Both the Ryzen 5950X system AND the 3930K system has an IB card in each system.
On the Ryzen 5950X system, when I try to boot from the thumb drive, it will produce the kernel panic as shown above.
Booting directly into the USB drive on the 5950X system produces this kernel panic:
Therefore; as a workaround, I am using my 3930K system (which also has an IB card in it as well), and I am using that to "jumpstart" my 5950X system.
Per the other thread, it was stated that CentOS 7.9.2009 is the
only supported CentOS version here, so that's what I'm using. So if I try to use that directly on the 5950X system WITHOUT first "jumpstarting" said 5950X system with my 3930K system, then it results in a kernel panic right away.
But if I want the 5950X system to work, then I would need to install CentOS 7.9.2009 using my 3930K system first, and then I can update the kernel (in order to get around the kernel panic issue 2.something seconds into the boot sequence) because if I do that, then at least I can get into the GNOME desktop environment. However, the moment that I update, for example, glibc to the latest, upon reboot is where/how/when you'd see the kernel panic as shown above.
Your help is greatly appreciated.
Thank you.
P.S. In regards to the "better drivers" comment from your reply above, I tried using the Mellanox MLNX_OFED_LINUX driver instead of the "inbox" Infiniband drivers that ships with the OS.
The problem is that with that, I would have to install the MLNX_OFED_LINUX drivers BEFORE I update the kernel using my 3930K system. And that works fine.
However, once I update the kernel, then even the Mellanox drivers doesn't work anymore because the kernel modules haven't been updated/rebuilt for that kernel (version) and as a result, `/etc/init.d/openibd restart` fails.
And if I try to execute `./mlnxofedinstall --add-kernel-support`, it fails to find the appropriate source for the kernel even if I install (when I update the kernel for example: `sudo yum install --enablerepo=elrepo-kernel kernel-lt kernel-lt-devel`), the Mellanox drivers doesn't know how or where to find the new kernel source properly and therefore; the rebuild of the kernel modules files, and therefore; `/etc/init.d/openibd restart` fails.
Therefore; either way:
1) Using the default 3.10.0-1160 kernel on 5950X results in a kernel panic which renders the entire system inoperative.
2) Updating the kernel using elrepo-kernel results in a kernel that's not supported here, but the "inbox" Infiniband drivers causes a kernel panic, and the Mellanox drivers can't rebuild the kernel modules for the updated kernel, which results in IB being inoperative in the system which is a no-go, full-stop barrier for me.
(This is being tested as a new compute node in a cluster, and therefore; getting the system, with IB operational is imperative.)
Thank you.
