Skip to content

memcg: Slow down swap allocation as the available space gets depleted

Jakub Kicinski cgroups linux-mm 7 versions – Applied

Status from this message: [merged] mm-automatically-penalize-tasks-with-high-swap-use.patch removed from -mm tree

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v6 Latest4 patches 2 repliesApplied 4/4
  2. v5 posting 2/24 patches 6 repliesSuperseded
  3. v5 posting 1/24 patches 1 replySuperseded
Earlier versions · 4
VersionPostedPatchesRepliesStatus
v44 patches5 repliesSuperseded
v33 patches5 repliesSuperseded
v23 patches11 repliesSuperseded
v13 patches31 repliesSuperseded

Patches in v6 All 4 applied

#Subject and filesSizeRepliesApplied
1/4mm: prepare for swap over-high accounting and penalty calculationmm/memcontrol.c+35 -270 repliesApplied
2/4mm: move penalty delay clamping out of calculate_high_delay()mm/memcontrol.c+8 -80 repliesApplied
3/4mm: move cgroup high memory limit setting into struct page_counterinclude/linux/memcontrol.h include/linux/page_counter.h mm/memcontrol.c+19 -110 repliesApplied
4/4mm: automatically penalize tasks with high swap useDocumentation/admin-guide/cgroup-v2.rst include/linux/memcontrol.h mm/memcontrol.c+102 -72 repliesApplied

Reviews trailers per patch

All 4 patches have a Reviewed-by or Acked-by.

PatchShakeel ButtJohannes Weiner
1/4R-b A-b
2/4R-b A-b
3/4R-b A-b
4/4A-b

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.

Read the v6 thread

Cover letter [PATCH mm v6 0/4] memcg: Slow down swap allocation as the available space gets depleted

Tejun describes the problem as follows:

When swap runs out, there's an abrupt change in system behavior - the anonymous memory suddenly becomes unmanageable which readily breaks any sort of memory isolation and can bring down the whole system. To avoid that, oomd [1] monitors free swap space and triggers kills when it drops below the specific threshold (e.g. 15%).

While this works, it's far from ideal:
- Depending on IO performance and total swap size, a given headroom might not be enough or too much.
- oomd has to monitor swap depletion in addition to the usual pressure metrics and it currently doesn't consider memory.swap.max.

Solve this by adapting parts of the approach that memory.high uses - slow down allocation as the resource gets depleted turning the depletion behavior from abrupt cliff one to gradual degradation observable through memory pressure metric.

[1] https://github.com/facebookincubator/oomd

v5: https://lore.kernel.org/linux-mm/20200521002411.3963032-1-kuba-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org/
v4: https://lore.kernel.org/linux-mm/20200519171938.3569605-1-kuba-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org/
v3: https://lore.kernel.org/linux-mm/20200515202027.3217470-1-kuba-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org/
v2: https://lore.kernel.org/linux-mm/20200511225516.2431921-1-kuba-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org/
v1: https://lore.kernel.org/linux-mm/20200417010617.927266-1-kuba-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org/

Jakub Kicinski (4):
mm: prepare for swap over-high accounting and penalty calculation
mm: move penalty delay clamping out of calculate_high_delay()
mm: move cgroup high memory limit setting into struct page_counter
mm: automatically penalize tasks with high swap use

 Documentation/admin-guide/cgroup-v2.rst |  20 +++
 include/linux/memcontrol.h              |   4 +-
 include/linux/page_counter.h            |   8 ++
 mm/memcontrol.c                         | 177 ++++++++++++++++++------
 4 files changed, 160 insertions(+), 49 deletions(-)
Signature · 1 line
2.25.4