makedumpfile fail to strip pages

[Martin Friedrich]

Hello,

we have built a new package using the latest commit (bd30437c6292) and are unable to reproduce the issue.
All dumps were correctly stripped and completed successfully.

We will monitor the situation and come back if we see further issues.
Thank you for your assistance.

Best,
Martin

> On 25. Sep 2024, at 03:36, HAGIO KAZUHITO(萩尾　一仁) <k-hagio-ab at nec.com> wrote:
> 
> On 2024/09/20 16:31, Martin Friedrich wrote:
>> Hello,
>> 
>> I try to identify the root-cause of an issue with kdump & makedumpfile which results in dumps to have a size of the system memory.
>> Kdump-tools/makedumpfile has been configured to strip most pages with the flags: " -c -d 31" and transfer the dump to an ssh target.
>> 
>> I'm unable to replicate the issue by crashing kernel using sysrq-triggers. As long the system crashed by sysrq-trigger, the dump has the correct size and completes successfully.
>> But as soon a system crashes for other reasons, dumps get huge and fail to transfer to the ssh target due to their size.
>> 
>> Our servers (x64/aarch64) are running Ubuntu 22.04 with a self built kernel 6.6.32 (config attached) and makedumpfile (backported 1.7.5-3 to Ubuntu 22.04).
>> 
>> Am I missing something within the kernel config which could cause this behavior?
> 
> hmm, I have not seen such a behavior.
> 
> - what happens with the latest (bd30437c6292) makedumpfile?
> - is it possible to get makedumpfile logs with "--message-level 31" on 
> sysrq crash and other reasons? (to compare the logs including stats)
> 
> Thanks,
> Kazu