I see

The Sysdig’s article says

As a last addendum to the story, this specific behavior of the dentry cache was fixed in more recent versions of the kernel. In particular, allocations in the kernel object pools are now tied to the limits of the memory cgroups that requested them, if any. This means that the trasher application wouldn’t have been able to generate all those invalid dentry objects to begin with. Unfortunately, the customer ran on a pretty old, but popular among enterprise customers, distribution and kernel version that didn’t contain those improvements (2.6.32 on RHEL 6).

I think the word “fixed” made me think it was a bug. I will try to fix the sentence in my article.

Thanks for pointing out!