This article describes how to generate a crashdump from an unresponsive XenServer 7.
If XenServer crashes, the running kernel can migrate into a special memory area which is based on kexec functions that loads a special kernel without the need for a server cold restart. This new kernel attempts to gather as much details about the crash as possible (memory dumps, process list) and save it into the /var/crash directory.
In some instances, XenServer hangs and does not reboot after a crash. In this situation, the server is unresponsive but still running and only a reboot of the server can resolve the issue. This could cause a situation where no crashdumps are available on the server, that is, no dump in the /var/crash directory and the xe host-crashdump-list contains no information about the crash.
In this situation, you can force a crashdump that saves information about the crash which can then be analyzed.
Requirements
To generate a crashdump file from an unresponsive XenServer 7 or later version, complete the following procedure:
1. If you are not using COM1 on your server, edit /etc/grub.cfg file as follows:
menuentry 'XenServer (Serial)' {
search --label --set root root-qeexyy
multiboot2 /boot/xen.gz com1=115200,8n1 console=com1,vga dom0_mem=2048M,max:2048M watchdog ucode=scan dom0_max_vcpus=8 crashkernel=128M@256M
module2 /boot/vmlinuz-4.4-xen root=LABEL=root-qeexyy ro nolvm hpet=disable console=tty0 xencons=hvc console=hvc0
module2 /boot/initrd-4.4-xen.img
}