Comment 15 for bug 1651609

Sending in my report now. My issue *maybe* is related to cgroups/docker. I don't think I saw this issue until I started trying to request hugepages from a docker container.

But from what I can see, some processes just... do not give up the pages after they exit. At least not in a way that the kernel can pick up on. Instead I end up with pages that are not available, I cannot free them, and I can find no evidence of them in /dev/hugepages, ipcs, or /proc/*/numa_maps. They are just ghosts!