[linux-azure] 19.04 regression of KDUMP

[Joseph Salisbury]

By default our test cases use DS1_v2 to test KDUMP, when we trigger a kernel panic, the VM enters into emergency mode, which can be seen on the serial console.

This issue can be resolved if the VM is resized from DS1_v2 (1 vcpus, 3.5 GiB memory) to DS12_v2 (4 vcpus, 28 GiB memory). However, Ubuntu 18.04 and Ubuntu 19.10 KDUMP work using a VM size of DS1_v2. This issue only happens when using Ubuntu 19.04.

The following appears on the 19.04 serial console when the issue occurs:

[ OK ] Reached target Local File Systems (Pre).
         Mounting /boot/efi...
         Starting File System Check…/cloud/azure_resource-part1...
[FAILED] Failed to mount /boot/efi.
See 'systemctl status boot-efi.mount' for details.
[DEPEND] Dependency failed for Local File Systems.
[ OK ] Started File System Check …sk/cloud/azure_resource-part1.
[ OK ] Started File System Check Daemon to report status.
         Starting GRUB failed boot detection...
         Starting Create final runt…dir for shutdown pivot root...
         Starting Tell Plymouth To Write Out Runtime Data...
[ OK ] Started Emergency Shell.
         Starting Load AppArmor profiles...
         Starting Create Volatile Files and Directories...
[ OK ] Stopped Dispatch Password …ts to Console Directory Watch.
[ OK ] Reached target Emergency Mode.

REPRO STEPS:

Create Ubuntu 19.04 using size Standard_DS1_v2.
Install kdump tools, reboot VM - apt update && apt install linux-crashdump, reboot VM, kdump-config show check the state is - ready to kdump.
Trigger kernel panic - sysctl -w kernel.sysrq=1 && echo c > /proc/sysrq-trigger

Expected behavior - VM reboot successfully and crash dump file and related subdirectories generated successfully.
Current behavior - The VM is inaccessible from SSH, it enters into emergency mode, have to press Ctrl + D in serial console, then VM enters into normal reboot process, crash dump file and related subdirectories generated successfully.

Update value for crashkernel to reserve the memory by updating the /etc/default/grub.d/kdump-tools.cfg file, run update-grub, reboot VM, dmesg | grep -i crashkernel check the memory is reserving successfully for crashkernel.

Then retest (trigger kernel panic) against different memory setting for crashkernel.

[Joseph Salisbury (jsalisbury)]

This bug only affects kdump on 19.04 and not the 5.0 kernel on 18.04. Since this is the case, I'll close this bug since 19.04 is EOL soon.