Configuring Kernel Crashdump

Our images are preconfigured with the kdump facility. During normal boot they will load a krash dump kernel into a reserved area of memory and, should the system crash, they will immediately kexec() into this kernel. When the initramfs detects that we're running a crash dump kernel, it will attempt to find a filesystem with the label HVCRASHDUMP. If found, it will mount this under /var/crash and copy the dump to it. Afterwards the system will perform a normal reboot.

To enable crashdump collection, you therefore need to:

Create a LUN accessible by a single blade of at least 2G (using web interface)
Create a single partition on the LUN (fdisk)
Create a filesystem on the partition (mkfs.ext4)
Label the filesystem HVCRASHDUMP (e2label)
To test crashdump:

echo c | sudo tee /proc/sys/sysrq-trigger