CentOS 8 Broken Installation

Home » CentOS » CentOS 8 Broken Installation
CentOS 24 Comments

Hello,

Is it a Plan to correct the Broken mdadm and or driver for Intel Chipset C602, my Server SuperMicro Board X9SRi-F have a Problem with the new C8 or RH8 ?

With C7 I can install CentOS with no Problem?

Have any a workaround or is it planed a Update for this Problem

Thanks for a answer,

24 thoughts on - CentOS 8 Broken Installation

  • Richard on CentOS-devel mailing list has similar problem but suggests it might be issue only with CentOS, he says he had to add the kernel line to make it work. Name of the tread is:
    [CentOS-devel] Booting CentOS 8 in a VM – can’t find the disk

    —
    Ljubomir Ljubojevic
    (Love is in the Air)
    PL Computers Serbia, Europe

    StarOS, Mikrotik and CentOS/RHEL/Linux consultant

  • Hello,

    Am Mittwoch, 2. Oktober 2019, 21:39:24 CEST schrieb Ljubomir Ljubojevic:

    NO, I downloaded and test RH 8 and RH Beta 8.1 with the same Problem?

    I mean the chipset C602 is not so old, but have a SCU inside this is the Problem?

  • I saw it uses Intel hdd controler so I thought it could be similar.

    Here is solution for instaling CentOS 8 on Dell T610 with SAS controler
    (copied from Mike Chan):

    Intro:

    ** Keyword: CentOS 8 LSI SAS2
    Trying CentOS 8 on an old machine (Dell T610) but the installer can’t find disks for installation.

    It turns out that the RH removed the support for LSI SAS2 controllers in the mpt3sas driver.

    I’ve found some articles about this, and is trying to install with elrepo-provided DUD (driver update disk).

    Will update the results and the procedure if succeeded.

    At least it can serves as a warning to those with said SAS controllers.

    Solution:

    Reference:
    https://elrepoproject.blogspot.com/2019/08/rhel-80-and-support-for-removed-adapters.html

    Here’s the procedure:
    1. prepare a install media and a flash drive, formatted with fat32, and copy the DUD iso file into the flash drive:
    https://elrepo.org/linux/dud/el8/x86_64/dd-mpt3sas-27.101.00.00-1.el8_0.elrepo.iso. You can also use tools to create a bootable install flash drive, and use it as the DUD drive. No need to process the iso; just place it in the drive.

    2. boot with the install media, and hit tab key in the boot menu, add
    ‘inst.dd’ to the command line, and then boot it.

    3. the installer will ask for a DUD before getting into the GUI. It will list the block devices it found, find the one with vfat – that’s your flash drive. Then it will list DUD isos it found on the drive, enter the number of the entry, then a “checkbox” will appear before the entry. Enter the number again so the “checkbox” is checked with a x. Now enter
    “c” to contiune.

    4. If everything works, you’ll have your disks visible to the installer. Install it as usual.

    5. The booting should has no problem, but DO NOT update kernel just yet. According to the reference page, the dracut tool in RHEL / CentOS 8.0 is bugged and will NOT include the extra driver to the updated initrd. This will make the new kernel unbootable. Setup network to download the patched dracut tools first, or copy it from other computer with flash drive:
    http://elrepo.org/people/akemi/testing/el8/dracut/
    Then install them with:
    rpm -Uvh (downloaded rpm files).

    6. Now you can update the kernel.

    In my case, I’m using the card in IT mode; the disks are not set into RAID with the card, so only the mpt3sas DUD is required. If you have a IR mode (RAID via the controller) configuration, you may need the megaraid-sas DUD too.

    Original post with solution:
    https://www.facebook.com/groups/CentOSproject/permalink/10157729112557728/?comment_id=10157729981242728

    —
    Ljubomir Ljubojevic
    (Love is in the Air)
    PL Computers Serbia, Europe

    StarOS, Mikrotik and CentOS/RHEL/Linux consultant

  • Am Mittwoch, 2. Oktober 2019, 21:32:24 CEST schrieb Jonathan Billings:

    OK, I can’t say I have a account ?

    This is a C602 Chipset this is not so old, but have a second Intel SCU
    Controller on Board and this is on RH / C8 the Problem!

    I tested it now with all RH 8 Based system and all have the Problem with this Controller?

    C7.7 is working and found the SCU ??

  • Since no one else chimed in, I did some googling, and it looks like someone has already filed a bug that looks pretty similar:

    https://bugzilla.redhat.com/show_bug.cgi?id=1710123 <https://bugzilla.redhat.com/show_bug.cgi?id=1710123>

    There is some advice there.

    It looks like the Intel C620 chipset is an intel “fake” RAID controller, and I’d recommend you turn that off and use software RAID in the future.

    If you are a Red Hat customer, I definitely suggest filing a support request.

    —
    Jonathan Billings

  • Jonathan Billings wrote:
    Oh, the Intel fakeRAID. I started reading about that maybe 8 years ago, and the general advice is “do not use”, use mdraid. Works for us.

    mark

  • This is not a very specific problem report. ‘a Problem’ doesn’t tell us anything. What exactly doesn’t work? since you mentioned mdadm, I’m guessing you’re having issues with SATA storage ?

    does the motherboard have the latest BIOS? are the SATA controller(s) in AHCI mode and not fake-RAID mode or legacy IDE emulation mode ?

    are you using the native PCH connected SATA ports, or the SCU (SAS) SATA
    ports?

    —
    -john r pierce
    recycling used bits in santa cruz

  • Hello,

    Am Mittwoch, 2. Oktober 2019, 22:36:37 CEST schrieb mark:

    Yes, it is a “fake” raid Controller bat is not possible to found the drives in Raid Mode or AHCI Mode ?

    This Board have 2 Controller!

  • If you provide the device ID pairing [xxxx:yyyy] for the one that is not working, it will become clear. Can you show us the output from:

    lspci -nn

    Is it this one?

    Serial Attached SCSI controller [0107]: Intel Corporation C602 chipset
    4-Port SATA Storage Control Unit [8086:1d6b]

    If so, the above device ID [8086:1d6b] is not supported in RHEL/CentOS
    8, unfortunately.

    Akemi

  • And if that is indeed your device, that was bad news. But you have good news too.

    Device [8086:1d6b] is supported by the isci kernel module. ELRepo has a DUD for this driver: dd-isci-1.2.0-1.el8.elrepo.iso

    Ljubomir provided detailed instructions on how to use ELRepo’s DUD
    isos earlier in this thread.

    Akemi

  • Hello,

    Am Donnerstag, 3. Oktober 2019, 01:10:58 CEST schrieb Akemi Yagi:
    wrote:

    ??
    Yes, I Have :-(

    07:00.0 Serial Attached SCSI controller [0107]: Intel Corporation C602 chipset
    4-Port SATA Storage Control Unit [8086:1d6b] (rev 06)?

    what is the Problem with this chipset and why it is no longer supported ??

    This Board is 4 year old and now???

    I mean it is in many server boards.

  • Hello,

    Am Donnerstag, 3. Oktober 2019, 01:41:59 CEST schrieb Akemi Yagi:
    wrote:

    OK, I will test it!

    Thanks fort the help and answers,

  • My impression is that you have it set to “RAID” instead of “AHCI” in BIOS. Switch in BIOS its operation to AHCI, you will see attached drives. Configure these drives as software RAID.

    As a matter of fact neither of fake RAID cards were ever supported by systems I know of (excluding MS Windows) as RAID cards. That is where jargon “fake RAID” widely used by Linux Folks comes from.

    System board manufacturers (motherboard is common jargon for system board for over 20 years) share their part in the spreading of fake RAID chips. Fake RAID chip is cheap (pun intended), so adding it to system board does not increase its cost much, but increases it apparent value in eyer of uninformed (I should say ignorant) mass consumer. It is probably time to call garbage (fake RAID) garbage and not expect from it to behave as real RAID (hardware RAID), and definitely not put blame on the system for garbage hardware being garbage hardware.

    Sorry about long post, I keep being upset by manufacturers who do this.

    Valeri

    ++++++++++++++++++++++++++++++++++++++++
    Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247
    ++++++++++++++++++++++++++++++++++++++++

  • The problem with the chipset is that the fake raid breaks under load for some subset of users and trying to debug the file corruptions can be impossible from the OS side. So you end up with a circle of finger pointing where the user points to the OS manufacturer. The OS
    manufacturer points to the hardware. The hardware will point to the user for using the hardware outside of its specific ‘recommendations’
    which you thought was a EULA. After a long set of time.. no one is happy and a lot of money/time was lost on something which can’t be fixed. So the hardware gets dropped from the OS vendor’s supported line to stop the circle. It doesn’t make the customer any happier.. but they weren’t going to be anyway.

    —
    Stephen J Smoogen.

  • På Thu, 3 Oct 2019 07:38:05 -0500
    Valeri Galtsev
    skrev:

    Of course he set it to RAID. He wouldn’t be able to boot his raid, if he set it to AHCI. If set to AHCI, it would require an extra disk to boot the system.

    Really ? You have really never heard of this small company called Intel, whos chipset have been able to boot Linux in RAID mode for a very long time ?

    You are right, a BIOS chip is really very cheap, and that is all that the chipset needs to be able to boot in RAID mode on the existing AHCI
    controllers. There is no raid “card” in these systems – aka no special extra raid cpu to control anything. All the RAID BIOS does is boot the system to load GRUB ( and maybe GRUB
    uses the same BIOS INT13H service to load the kernel). After that, kernel drivers take over – and that is MDADM doing the rest of the job, as Intel RAID bios is MDADM compatible.

    Call it FAKE raid, call it BIOS raid or whatever – it is in fact just a Linux software mdadm raid – with the added ability to boot directly from the BIOS. I don’t see any reason to call this a bad solution or unstable in any way more than what mdadm is.

    I have myself a small desktop/server system here running for almost 7
    years according to SMART on my disks – booting from such a Intel Z77
    chipset to a mdadm raid 5.

    I do run Fedora on this system as RH most likely have removed my chipset from CentOS 8 too.

    Allan.

  • The SCU in question is a *SAS* controller, so there’s no AHCI mode, thats for native SATA stuff, not for SAS stuff.

  • You are wrong. Mdraid (mdadm is only a tool for mdraid) is SOFTWARE
    RAID, to create RAID system of regular or AHCI chipset. No part of mdraid uses BIOS, it is all in kernel/driver.

    What you described is actually “dmraid” (discover, configure and activate software (ATA)RAID ). Dmraid is just an interface for fakeRAID
    BIOS to work, and bugs in fakeRAID BIOSes (for which firmare updates barely exist) are the reason (some) fakeRAID chipsets are removed from kernel.

    —
    Ljubomir Ljubojevic
    (Love is in the Air)
    PL Computers Serbia, Europe

    StarOS, Mikrotik and CentOS/RHEL/Linux consultant

  • Hello,

    Am Mittwoch, 2. Oktober 2019, 22:22:09 CEST schrieb Ljubomir Ljubojevic:

    on this morning I tested this with a DUD CD from Elrepo. I like to tell all it is working again :-).

    I found my drives again….

    Many many thanks for all helped People,

  • How to see if your HDD controler is not supported:

    You can check by running the lspci command (lspci -nn) and compare it with Red Hats list of removed ID’s:
    https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html/considerations_in_adopting_rhel_8/hardware-enablement_considerations-in-adopting-rhel-8#removed-adapters_hardware-enablement

    Here is the list of all (so far created) the DUD’s for EL8:
    https://elrepo.org/linux/dud/el8/x86_64/

    And here is ElRepo page http://elrepo.org/tiki/DeviceIDs that shows all the drivers ElRepo has the drivers for, and can create DUD ISO files if asked via ElRepo bugzila:
    https://elrepo.org/bugs/main_page.php

    —
    Ljubomir Ljubojevic
    (Love is in the Air)
    PL Computers Serbia, Europe

    StarOS, Mikrotik and CentOS/RHEL/Linux consultant