NFS Kdump for iSCSI boot image system is not working

Issue

In a diskless HPC nodes context, the nodes are booted using an iSCSI root device. It is also needed to set up the kernel crash dump using an NFS server.

When a machine, booted via iSCSI, crashes, the NFS enabled initrd kdump will fail to mount the root device with a message “Ignoring netroot argument for NFS” and “/dev/mapper/mpathb1 does not exist”.

This is the content of etc/kdump.conf :

nfs dump.srv.totox:/var/dump
path /
core_collector makedumpfile -l --message-level 1 -d 31
default shell

Extract from the kdump initrd sequence :

Command line: ro initrd=initramfs-dkls-overlayfs.img initrd=dkls_custom_all.cpio initrd=dkls_custom_rome13.cpio root=/dev/mapper/mpathb1 netroot=iscsi:diskless1.srv.totox::iqn.2015-08.fr.totox:rome13-target netroot=iscsi:diskless2.srv.totox::iqn.2015-08.fr.totox:rome13-target iscsi_initiator=iqn.2015-08.fr.totox:rome13-initiator LANG=en_US.UTF-8 console=tty0 console=ttyS1,115200 text=1 vconsole.font=latarcyrheb-sun16 LANG=en_US.UTF-8 ip=eno1:dhcp rootflags=noload intel_pstate=disable rd.iscsi.param=node.session.initial_login_retry_max=0 rd.iscsi.param=node.conn[0].timeo.login_timeout=1 dkls.customdir=/ dkls.overlayfs selinux=0 irqpoll nr_cpus=1 reset_devices cgroup_disable=memory mce=off numa=off udev.children-max=2 panic=10 rootflags=nofail acpi_no_memhotplug disable_cpu_apicid=0 elfcorehdr=916872K

dracut-cmdline[79]: Warning: Ignoring netroot argument for NFS
…
dracut-initqueue[265]: Warning: Could not boot.
dracut-initqueue[265]: Warning: /dev/mapper/mpathb1 does not exist