I’m trying to install alma linux 8.5 (edited: 8.6 not 8.5) a dell precision 3571 with pxe boot and kickstart. The setup works fine with a DELL Precision 3520 but on this new laptop it hangs on:
“Started cancel waiting for multipath siblings of nvme0n1”
I saw some report about this problem described as a known bug, but no working workaround found. I’m stuck with this, unable to install this laptop…
Any help is wellcome
AlmaLinux is based on sources of RHEL that Red Hat releases. RH has released nothing for RHEL 8.5 since they released RHEL 8.6. Therefore, the only option is to try install with AlmaLinux 8.6. Either it has the bug fixed, or RHEL 8 still has the bug and RH should be told that. (Although, RH did just release beta of RHEL 8.7 which again should have new fixes and features.)
thanks for your answer. Ideed there is a mistake in my question, I am using Alma Linux 8.6, not 8.5. I’m sorry.
Is there a way to reach this next version of alma linux 8.7, even in beta version ? Or an older one ? I can only dowload 8.6 or 9.0 on almalinux.org
Booting on an USB key works but if I add inst.repo and inst.ks parameters in the grub command line I fall back to the waiting for multipath problem.
I also found this bug report 0000209: PXE Installer for 8.5 stuck at multipath - AlmaLinux Bug Tracker and the reporter (trex-thedog ) says that modifying initrd.img can help but do not provides any details about this modification. I was unable to proceed.
I have some additional information about this problem. The following screen shot shows the freezing situation:
But my laptop is using a dock station (via usb-c) to connect to the network and the BIOS is passing the laptop mac address through it. In the dhcp server log an IP is provided for the PXE boot. But when the install process hangs, I can see dhcp requests with the mac address of the dock station and no more with the allowed mac address of the laptop.
And this make the install process freeze as pluging the ethernet cable directly in the laptop RJ45 connector is working (after modifying the dhcp setup of course to manage this mac address).
Entry in RHEL bugzilla 1916168 – [RHEL8.4] system hung at Started cancel waiting for multipath siblings of x
seems to explain the root issue.
The initrd.img is a compressed file that contains a filesystem. There are tools* for extracting files from it and for creating new img from a directory tree.
*At least there were when I used them for my CentOS 5 setup. Alas, I can’t remember what they were.
this seams more related to the hardware in my mind. If I use my U2723QE DELL screen acting as a dock station, the DHCP problem of requesting an IP with the mac address of the dock at install time exist, even if the PXE boot request is based on the laptop mac address.
If I use a simple WD15 DELL dock the problem does not exist.
So with a network setup failure, the kickstart install cannot proceed anymore.
Yes, the network …
AFAIK, in boot:
- PXE hardware requests network config from DHCP server, including “next-server”
- PXE hardware loads bootloader (syslinux or grub) from “next-server” with TFTP
- Bootloader loads kernel from “next-server” with TFTP
- Kernel starts loads (from network) and starts installer
- Within installer you can reconfigure network
I have always failed step 3 when I had tagged VLAN, bootloader failed to use network connection that PXE got.
On step 4 the kernel could use connection data from PXE, kernel command-line parameters, or DHCP. DHCP is the default, the other require kernel command-line parameters (that bootloader can hand out).
If the issue is merely that DHCP server receives requests from multiple MAC addresses, then it could be set to hand out unique address for each MAC. You probably want to clean them up, when system is up and no longer uses both MACs.