Luigi Semenzato writes:
+One possible workaround (besides terminating enough processes) is to force
+excess anonymous pages out to swap before hibernating. This can be achieved
+with memcgroups, by lowering memory usage limits with ``echo <new limit> >
s/memcgroups/memory cgroups/
I also wouldn't say "excess", it implies that the cold(er) pages are
unnecessary, which may not be true.
+/dev/cgroup/memory/<group>/memory.mem.usage_in_bytes``. However, the latter
+operation is not guaranteed to succeed.
This is cgroup v1, which is frozen and in the process of being deprecated. We
should at least mention memory.high or similar.
Also, /dev/cgroup? /sys/fs/cgroup has been more or less standardised for quite
a while now...
(I also wonder whether the need for this in general is something the user
shouldn't have to care about and we should improve management of in-kernel, but
I know very little about hibernation so I'll leave that alone.)
