[Arm-dev] kexec: failed to load kdump kernel

I have 40 brand new Cavium R270-T60-V6 systems (T32 BIOS) with CentOS
7.3.1611 installed.  Five of them are failing at boot with:

 	kdumpctl[9698]: Could not find a free area of memory of 0x10000 bytes...
 	kdumpctl[9698]: locate_hole failed
 	kdumpctl[9698]: kexec: failed to load kdump kernel
 	kdumpctl[9698]: Starting kdump: [FAILED]

The systems are identical in every way (Memory, CPU, disks, firmware,
network, location....).

A comparison of their "dmesg" output shows no smoking gun, but I do not
know if differences like this are significant:

<  efi:  ESRT=0x13ffe1b0e98  SMBIOS 3.0=0xfffeb000  ACPI 2.0=0x13feaba0000 
---
>  efi:  ESRT=0x13ffe1b0e98  SMBIOS 3.0=0xfffeb000  ACPI 2.0=0x13feab9f000

<  Memory: 533473024K/536820736K available (7232K kernel code, 1406K rwdata, 3456K rodata, 1152K init, 6589K bss, 2823424K reserved, 524288K cma-reserved)
---
>  Memory: 533473152K/536820736K available (7232K kernel code, 1406K rwdata, 3456K rodata, 1152K init, 6589K bss, 2823296K reserved, 524288K cma-reserved)

Any tips on where to look next are appreciated.

(Also, if there is a more appropriate distro for this question, please advise).

Peace.
--
Matt McLean
ARC-TS Operations
mattmc at umich.edu