Why Does the OS Fail to Respond When kdump Occurs on a Linux ECS?

Symptom

When kdump occurs on a Xen Linux ECS, the OS fails to respond and cannot be automatically recovered. For example, if you run the echo c>/proc/sysrq-trigger command to trigger kdump, this fault occurs.

Figure 1 Triggering kdump
Click to enlarge

Generally, kdump is disabled for public images. This issue does not occur on the ECSs created using public images.

Possible Causes

Certain Linux kernel versions are incompatible with Xen virtualization.
If kdump is enabled in the ECS with the kernel not supporting soft_rest, the ECS stops responding during dump.

Solution

Method 1: Disable kdump.

CentOS 7.5 is used as an example in the following.

Forcibly restart the ECS.
1. Log in to management console.
2. Under Computing, choose Elastic Cloud Server.
3. In the ECS list, select the target ECS and click Restart.
4. Select Forcibly restart the preceding ECSs or Forcibly stop the preceding ECSs.
5. Click OK.
Disable kdump.
1. Log in to the forcibly restarted ECS as user root.
2. Run the following command to disable kdump:
  service kdump stop

Method 2:

If the target ECS supports the crash_kexec_post_notifiers function, add the function to the ECS startup configuration file (menu.lst or grub.cfg). To do so, perform the following operations:

Run the following command to check whether the ECS supports the crash_kexec_post_notifiers function:
cat /proc/kallsyms |grep crash_kexec_post_notifiers

Figure 2 Support for the crash_kexec_post_notifiers function
- If yes, go to step 2.
- If no, use method 1.
Add the crash_kexec_post_notifiers function to the startup configuration file (menu.lst or grub.cfg).
Take menu.lst as an example.
1. Run the following command to open the menu.lst file:
  vi /boot/grub/menu.lst
2. Add the crash_kexec_post_notifiers function to the startup item.
  Figure 3 Editing the menu.lst file
3. Run the following command to restart the ECS for the modification to take effect:
  reboot