Description and cautions

This article describes how to configure dump for capturing memory dumps, including application memory.

Details

1. Configure kdump
   • Altlinux
     There is no kdump-tools package in the default repository, so it has to be downloaded from the sisyphus repository:
     • Go to https://packages.altlinux.org/en/sisyphus/srpms/kdump-tools/ 
     • In List of rpms provided by this srpm select the kdump-tools package for the required architecture (can be checked by running uname -m)
     • Download the package from the Download link

     • Install it by running 

       apt-get update && apt-get install <path to the downloaded rpm>
       

     • After that, follow the Debian instruction from Edit /etc/default/kdump-tools step

   • Red Hat based distributions (tested on Fedora 38, Rocky Linux 9, Red OS)

     • Install kexec-tools

       sudo dnf install kexec-tools

     • Edit /etc/kdump.conf. In the configuration file edit the core_collector setting: option -d should be set to 17 instead of 31
     • Edit /etc/default/grub. Edit GRUB_CMDLINE_LINUX, add crashkernel=256M to reserve enough RAM for the dump kernel to run, and nmi_watchdog=1, to capture a dump in case of a system hang

     • Run 

       sudo grub2-mkconfig -o /boot/grub2/grub.cfg

     • Reboot

     • Enable kdump service 

       sudo systemctl enable --now kdump.service
       

   • Debian based distributions (tested on Debian, Astra CE, Alt Linux)

     • Install kdump-tools 

       sudo apt update && sudo apt install kdump-tools -y
       

     • Edit /etc/default/kdump-tools. In the configuration file edit the MAKEDUMP_ARGS variable: option -d should be set to 17 instead of 31
     • Configure the bootloader
       • In /etc/default/grub edit GRUB_CMDLINE_LINUX_DEFAULT, add nmi_watchdog=1 to capture a dump in case of a system hang
       • In /etc/default/grub.d/kdump-tools.cfg change crashkernel value to 384M-:256M (default is 384M-:128M)
         Expected result: GRUB_CMDLINE_LINUX_DEFAULT="$GRUB_CMDLINE_LINUX_DEFAULT crashkernel=384M-:256M"
       • ave and run sudo update-grub

   • SUSE Linux

     • Install kdump

       sudo zypper in kdump kexec-tools

     • Edit /etc/sysconfig/kdump
       Change KDUMP_DUMPLEVEL variable to 17

     • Edit /etc/default/grub
       Edit GRUB_CMDLINE_LINUX_DEFAULT, add crashkernel=256M to reserve enough RAM for the dump kernel to run, and nmi_watchdog=1, to capture a dump in case of a system hang

     • Update the bootloader configuration

       sudo grub2-mkconfig -o /boot/grub2/grub.cfg

     • Reboot

     • Enable kdump service

       sudo systemctl enable --now kdump.service

2. Configure SysRq dump trigger
   To enable SysRq trigger, these key combinations 'kernel.sysrq = 8'(without quotes) has to be added  to /etc/sysctl.conf.

    

   In SUSE the value of kernel.sysrq has to be changed in /usr/lib/sysctl.d/50-default.conf instead of /etc/sysctl.conf

3. Reboot or run 

   sudo sysctl --system

4. After the set up above is complete, to manually trigger a dump press Alt+SysRq, Alt+C. Alternatively:

   echo 8 | sudo tee /proc/sys/kernel/sysrq

   (Command above is only needed if kernel.sysrq is not set in /etc/sysctl.conf)

   echo c | sudo tee /proc/sysrq-trigger

    

5. Location of the dump files may vary between different Linux versions, it is configurable in the kdump configuration file. In Debian based distributions it is set by KDUMP_COREDIR variable. In Red Hat based distributions it is set by the path setting, generally the default location is /var/crash. Make sure that the dump folder has enough free space for the dump to be written. You may search by filemask: vmcore.

Related Information